Ebook An Introduction To Artificial Intelligence in Education Yu Online PDF All Chapter

An Introduction to Artificial Intelligence
in Education Yu
Visit to download the full and correct content document:
https://ebookmeta.com/product/an-introduction-to-artificial-intelligence-in-education-y
u/
More products digital (pdf, epub, mobi) instant
download maybe you interests ...
Artificial Intelligence With an Introduction to Machine

Learning Second Edition Richard E Neapolitan Xia Jiang
https://ebookmeta.com/product/artificial-intelligence-with-an-
introduction-to-machine-learning-second-edition-richard-e-
neapolitan-xia-jiang/
Introduction to Algorithmic Marketing Artificial

Intelligence for Marketing Operations Ilya Katsov
https://ebookmeta.com/product/introduction-to-algorithmic-
marketing-artificial-intelligence-for-marketing-operations-ilya-
katsov/
Artificial Intelligence in STEM Education : The

Paradigmatic Shifts in Research, Education, and
Technology 1st Edition Fan Ouyan
https://ebookmeta.com/product/artificial-intelligence-in-stem-
education-the-paradigmatic-shifts-in-research-education-and-
technology-1st-edition-fan-ouyan/
A Machine Learning Artificial Intelligence Approach to

Institutional Effectiveness in Higher Education 1st
Edition John N Moye
https://ebookmeta.com/product/a-machine-learning-artificial-
intelligence-approach-to-institutional-effectiveness-in-higher-
education-1st-edition-john-n-moye/
Artificial Intelligence in Higher Education: A
Practical Approach 1st Edition Prathamesh Padmakar
Churi
https://ebookmeta.com/product/artificial-intelligence-in-higher-
education-a-practical-approach-1st-edition-prathamesh-padmakar-
churi/
An Introduction to Education 1st Edition Hasan Arslan

(Editor)
https://ebookmeta.com/product/an-introduction-to-education-1st-
edition-hasan-arslan-editor/
Neural Assemblies An Alternative Approach to Classical

Artificial Intelligence Günther Palm
https://ebookmeta.com/product/neural-assemblies-an-alternative-
approach-to-classical-artificial-intelligence-gunther-palm/
Cognitive Electronic Warfare: An Artificial

Intelligence Approach Haigh
https://ebookmeta.com/product/cognitive-electronic-warfare-an-
artificial-intelligence-approach-haigh/
Swarm Intelligence An Approach from Natural to

Artificial 1st Edition Kuldeep Singh Kaswan
https://ebookmeta.com/product/swarm-intelligence-an-approach-
from-natural-to-artificial-1st-edition-kuldeep-singh-kaswan-2/
Bridging Human and Machine:
Future Education with Intelligence
Shengquan Yu
Yu Lu
An Introduction to
Artificial Intelligence
in Education
Bridging Human and Machine: Future
Education with Intelligence
Series Editors
Shengquan Yu, School of Educational Technology, Beijing Normal University,
Beijing, China
Ig Ibert Bittencourt , Computing Institute, Federal University of Alagoas,
Maceió, Brazil
This book series will gather researchers from AI to brain science, scientists from data
science to cognitive neuroscience; experts from academia to industry, and specifi-
cally pay attentions on the new frontiers and technologies that would transform
future education. This series intends to empower readers a deeper understanding
on future education from the perspectives of both enabling technology and learning
science. Topics are covered but not limited to the cutting-edge and multidisciplinary
research, development, and practice among the fields in education and technology.
For example, some books would address what and how AI techniques can be used for
enhancing learning and education system. Some books would seek an explanation
on how brain science can connect human brain’s physical activities and cognitive
process. Moreover, this book series would provide policy-makers and educational
stakeholders more quality and equity education, and encourage government and
private sectors to invest more resources in the development of educational tools and
facilities with intelligence.
More information about this series at https://link.springer.com/bookseries/16427

Shengquan Yu · Yu Lu
An Introduction to Artificial
Intelligence in Education
Shengquan Yu Yu Lu
Advanced Innovation Center for Future Advanced Innovation Center for Future
Education, Faculty of Education Education, Faculty of Education
Beijing Normal University Beijing Normal University
Beijing, China Beijing, China
ISSN 2662-5342 ISSN 2662-5350 (electronic)

Bridging Human and Machine: Future Education with Intelligence
ISBN 978-981-16-2769-9 ISBN 978-981-16-2770-5 (eBook)
https://doi.org/10.1007/978-981-16-2770-5
© Springer Nature Singapore Pte Ltd. 2021

This work is subject to copyright. All rights are reserved by the Publisher, whether the whole or part of
the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation,
broadcasting, reproduction on microfilms or in any other physical way, and transmission or information
storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar methodology
now known or hereafter developed.
The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication
does not imply, even in the absence of a specific statement, that such names are exempt from the relevant
protective laws and regulations and therefore free for general use.
The publisher, the authors and the editors are safe to assume that the advice and information in this book
are believed to be true and accurate at the date of publication. Neither the publisher nor the authors or
the editors give a warranty, expressed or implied, with respect to the material contained herein or for any
errors or omissions that may have been made. The publisher remains neutral with regard to jurisdictional
claims in published maps and institutional affiliations.
This Springer imprint is published by the registered company Springer Nature Singapore Pte Ltd.
The registered company address is: 152 Beach Road, #21-01/04 Gateway East, Singapore 189721,
Singapore
Preface
How to transform education with artificial intelligence (AI) has been one of the
focal research directions and key work of Advanced Innovation Center for Future
Education of Beijing Normal University in China. The book is produced from our
ongoing work in this area. The purposes of this book are to review existing solutions
of AI in education and to build bridges in schools, industries, and research institutions
so as to jointly explore the possibility of transforming education with AI.
To power education with AI, we are supposed to focus on both specific technolo-
gies and the real problems in educational practices. Generally speaking, there are
three core components in applying AI in education, including computational power,
intelligent algorithms, and real-world scenarios. Among them, knowledge of real-
world application scenarios is critical to the success of AI in education. Therefore,
this book aims to understand AI in education in terms of application scenarios and
pressing issues in educational practices.
Focal themes of AI in education should not be confined only to the intelligent
supports to students’ learning, as well as to the test and assessment of students’
performance. The other aspects of equal importance include the support to students’
healthy growth and overall development, the support to improve teachers’ work
efficiency, and the support to education administrators. Reforms in education can
only occur when AI is seamlessly integrated into the vital services of education.
This book is edited by the research team of Advanced Innovation Center for Future
Education of Beijing Normal University. We invited experts from fields of computer
science, education, psychology, cognitive neuroscience, and so on to form an edito-
rial board and expert committee. After a whole year’s joint efforts and several rounds
of revision, this book finally comes into being. The book is comprised of nine chap-
ters. Chapter 1 introduces the impact of AI on education and related background. In
Chap. 2, we review primary technologies of AI and general technical solutions in
educational contexts. Chapters 3–7 show how AI is applied in education in terms of
smart learning environments, AI-facilitated learning process, AI-enhanced learning
v
vi Preface
assessment, intelligent support for teachers, and AI-supported education manage-

ment and services. Chapter 8 summarizes major themes and trends of AI in educa-
tion. Critical issues and challenges in applying AI in education are discussed in the
final chapter.
In order to make AI in education more accessible to practitioners, this book
involves a large number of real-life cases. In the early phase of this work, we gathered
a collection of cases from schools, research institutions, and enterprises so as to
understand the practices being carried out by different sectors of the community in
China. Cases involved in the book represent the specific scenarios and directions of
solving educational problems with AI. On behalf of the editorial board, I am grateful
to all the schools, organizations, and companies who made this book possible.
The technologies of AI are developing rapidly. However, reforms and changes in
the field of education are taking place steadily. It is impossible for us to cover in this
book all aspects of AI in education. Your advice or suggestions, in this regard, will
be much appreciated and welcomed. For any questions and advice, you can always
reach me at [email protected]. Thank you very much!
Beijing, China Shengquan Yu

Contents
1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.1 Cognition and Learning in the Age of AI . . . . . . . . . . . . . . . . . . . . . . . 2
1.2 A Brief History of AI in Education . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.3 Typical Areas of AI in Education . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
2 An Overview of AI . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
2.1 Basic Approaches in AI . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
2.2 Major AI Technologies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
2.3 Typical Open-Source AI Projects and Cloud Services . . . . . . . . . . . . 22
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
3 Intelligent Learning Environments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
3.1 Smart Campus . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
3.2 Campus Safety Detection and Early Alarming . . . . . . . . . . . . . . . . . . 35
3.3 Smart Classrooms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
3.4 Smart Library . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
3.5 Intelligent Writing System . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
4 Intelligent Learning Processes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
4.1.1 Intelligent Mapping and Visualizing of Knowledge
and Competency . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
4.1.2 Intelligent Diagnoses and Recommendations . . . . . . . . . . . . . 58
4.1.3 Monitoring Learning Load . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
4.1.4 Virtual Environment for Inquiry-Based Learning . . . . . . . . . 66
4.1.5 Intelligent Domain-Specific Tools . . . . . . . . . . . . . . . . . . . . . . 71
4.1.6 Intelligently Assisted Learning Based on Brain Science . . . . 80
4.1.7 Intelligent Robotics and Toys . . . . . . . . . . . . . . . . . . . . . . . . . . 85
4.1.8 Intelligent Assistant in Special Education . . . . . . . . . . . . . . . . 92
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99
vii
viii Contents
5 Intelligent Teacher Assistant . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101

5.1 Intelligent Tutoring . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102
5.2 Intelligent Question Answering . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 108
5.3 Intelligent Test Generation and Evaluation . . . . . . . . . . . . . . . . . . . . . 110
5.4 Generation of Report for Education . . . . . . . . . . . . . . . . . . . . . . . . . . . 115
5.5 Personalized Assignment and Dialogue System . . . . . . . . . . . . . . . . . 117
5.6 Automated Assessment of Speaking . . . . . . . . . . . . . . . . . . . . . . . . . . . 119
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123
6 Intelligent Educational Evaluation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125
6.1 The Diagnosis and Evaluation of Problem-Solving Skills . . . . . . . . . 126
6.2 Mental Health Monitoring, Early Warning, and Intervention . . . . . . 131
6.3 Physical Health Monitoring and Improvement . . . . . . . . . . . . . . . . . . 135
6.4 Smart Class Evaluation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 140
7 Intelligent Educational Management and Service . . . . . . . . . . . . . . . . . . 141
7.1 AI to Promote Education Equity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 142
7.2 Big Data to Support Intelligent Decision-Making in Education . . . . 146
7.3 Intelligent Technologies to Improve the Quality of Regional
Teaching . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 151
7.4 Dynamic Monitoring of Education Quality . . . . . . . . . . . . . . . . . . . . . 153
7.5 Customized Education Services . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 157
7.6 Intelligent Medical Assistants at School . . . . . . . . . . . . . . . . . . . . . . . 160
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 168
8 Frontiers of AI in Education . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 169
8.1 Educational Knowledge Graph . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 170
8.2 Cognitive Diagnosis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 172
8.3 Student Modeling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 174
8.4 Educational Question Representation . . . . . . . . . . . . . . . . . . . . . . . . . . 176
8.5 Machine Reading Comprehension and Scoring . . . . . . . . . . . . . . . . . 178
8.6 Intelligent Tutoring System . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 180
8.7 Neuroscience and Education . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 182
8.8 Human–Machine Integration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 183
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 184
9 Prospects and Reflections: Looking into the Future . . . . . . . . . . . . . . . . 189
9.1 Prospects of AI in Education . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 189
9.2 Reflections on AI in Education . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 192
9.3 Suggestions for Promoting AI in Education . . . . . . . . . . . . . . . . . . . . 194
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 198
Chapter 1
Introduction
The concept of artificial intelligence (AI) came into being in 1956. For over half
a century, the development of AI technologies and disciplinary applications has
experienced many highs and lows. Since 2006, machine learning, especially deep
learning models, has achieved great success in the fields of computer vision and
speech recognition. As the accuracy of recognition has dramatically improved, AI
has widely drawn people’s attention in both academia and industry [1].
In July 2017, the State Council of People’s Republic of China officially issued
the Next Generation of AI Development Plan and identified a three-step goal. Under
the joint drive of new theories and technologies, such as mobile Internet, big data,
supercomputing, sensor network, brain science, and strong demand for economic and
social development, the development of AI has accelerated. So far, AI has become one
of the new foci of international competition and an engine of economic development.
The development of AI should take the urgent needs of education, medical care,
and safety and security into consideration and provide diverse and personalized
high-quality services to the public. The concept of intelligent education has been
put forward in order to use AI technology to accelerate the reform of education,
including changes in the talents development model and teaching methods to build
a new education system [2].
Around the world, many countries have released relevant policies and reports to
deepen the development and application of AI technology. For instance, in 2016,
the USA issued Preparing for the Future of Artificial Intelligence [3] and National
Artificial Intelligence Research and Development Strategic Plan [4] in succession.
Besides, Stanford University released Artificial Intelligence and Life in 2030 [5] as
the first product of the ongoing project—“One Hundred Year Study on Artificial Intel-
ligence (AI100).” Pearson, in collaboration with UCL Knowledge Lab, co-published
Intelligence Unleashed: An Argument for AI in Education [6]. These policy docu-
ments have brought about significant global impact. Moreover, AI was listed as one
of the core technologies that would transform teaching and learning in the next five
© Springer Nature Singapore Pte Ltd. 2021 1

S. Yu and Y. Lu, An Introduction to Artificial Intelligence in Education,
Bridging Human and Machine: Future Education with Intelligence,
https://doi.org/10.1007/978-981-16-2770-5_1
2 1 Introduction
years in the 2016 Horizon Report (basic education edition) [7] as well as the 2017
Horizon Report (basic education edition and higher education edition) [8, 9].
1.1 Cognition and Learning in the Age of AI
Intelligence is the intellectual capacity of humans, including perception, memory,

reasoning, behavior, and language [10], while AI refers to the intelligence that
is implemented manually on machines (including computers) or that people use
machines to simulate the intelligence of humans and other organisms [11]. As we
all know, human beings have limits in physical strength, energy, and intelligence.
AI empowers human’s natural intelligence to go beyond limits to a certain degree.
Instead of replacing humans in certain areas, AI serves to extend and expand human
capacity [12].
In the Information Age, distributed cognition has become the fundamental way
of understanding humans’ mind and examining how human beings adapt to the
complex society [13]. Distributed cognition holds that the nature of human cognition
is distributed in itself. It is also believed that cognition occurs not only in our minds
but also via the process of interaction between humans and tools. As the gaps between
individual’s learning time, varied learning abilities, and the knowledge explosion of
modern society grow, we increasingly rely on the distributed cognition of people
and smart devices as well as on collaboration between humans and machines. In this
way, we can adjust ourselves to the increasingly sophisticated world [14].
The integration of humans and machines has been the primary way of knowing
the world. As is often the case, nowadays, the completion of many tasks requires
collaboration between humans and machines. Both the human brain and computers
are tools for information processing. The human brain excels in making decisions
based on imprecise and qualitative information. Computers have the faculty to do fast
and accurate quantitative calculation. The integration of the two thus complements
each other. With the development of technology, machines can better perceive the
emotion and behavior of people during the process of cognition and learning to offer
more personalized learning support and service. AI not only could reach the great
achievement of human wisdom but could also enhance human wisdom by providing
feedback on human’s performance [15].
1.2 A Brief History of AI in Education
AI technologies have a long history of supporting effective learning. The Intelli-

gent Tutoring System (ITS) first originated from the teaching machine designed
by behavioral psychologist Burrhus Frederic Skinner in the 1950s [16]. Since the
1970s, intelligent teaching systems have emerged in various disciplines, such as
the Socratic dialogue-based geography teaching system SCHOLAR and SOPHIE
1.2 A Brief History of AI in Education 3
and BUGGY designed for procedure-based teaching. Except for intelligent teaching
systems, projects of the large-scale knowledge base based on knowledge represen-
tation technology are also one aspect of applying AI in education. Early knowledge
bases were generally created by the government or the military (such as CYC in the
USA, YAGO in Germany). As per the development of the Internet, individuals began
to contribute to the knowledge base spontaneously, with Wikipedia being a typical
case [17]. Recently, the advancement of natural language processing technology
has promoted the human–computer interaction of intelligent teaching systems. For
example, the online robot teaching assistant created by Georgia Tech University with
the help of IBM Watson AI system has reached a reliability rate of 97% [18].
Meanwhile, AI has also been widely used for intelligent assessment, especially
in the field of foreign language learning (automatic evaluation of pronunciation,
essay scoring, etc.). The development of machine learning technologies can make
an accurate prediction of learner’s learning experience and offer a corresponding
recommendation of support or resources according to the analysis of data collected
during the process of learning. This is also how adaptive learning is achieved. More-
over, it can also collect data about learner’s body movements and facial expressions
during the learning process. Analyses of these data help to identify learner’s status,
which also serves as an input for teachers to improve their teaching.
The next generation of AI technologies has come on the stage. They will impact
education in all its forms. For example, personalized learning will get great support
from big data intelligence. Besides, cross-media intelligence will improve learner’s
efficiency and interest. Lifelong learning will benefit from intelligent support as well.
AI technologies will also transform the construction of libraries [19]. Application of
AI in education enables an accurate representation of various sources of knowledge.
If AI can accurately understand students’ learning data, teachers, facilitated by AI
technologies, can offer students appropriate support for learning, suitable learning
content and activities as well as personalized learning services in the future. Mean-
while, AI technologies connect formal and informal learning environments, which
further make ubiquitous learning possible. With intelligent knowledge network and
social network, which are integrated by ubiquitous terminals, we can engage in
lifelong learning anytime and anywhere [20]. In 2013, Woolf and colleagues put
forward five grand challenges that AI in education need to address: (1) offering each
learner with virtual mentors; (2) helping learners to master twenty-first-century skills
and support self-monitoring and self-assessment; (3) analyzing interaction data to
support learning; (4) creating universal access to global classrooms to foster global
connection; and (5) promoting lifelong and life-wide learning [21].
AI in education involves the application of AI technologies in education (learning),
the construction of educational contexts, and the reorganization of major components
of education or the reconstruction of educational processes [22]. Meanwhile, it is
believed that educational AI (EAI) is a new filed based on the integration of AI
technologies into learning sciences. EAI aims to understand how learning occurs and
examine how learning is impacted by various external factors (e.g., social economy
situations, physical environment, science, and technology) with AI technologies, and
to use this knowledge to support effective learning [23].
4 1 Introduction
To sum up, current research and practice of AI in education focus on the following
two aspects: (1) the application of AI technologies in educational practices (including
support for cognition, learning, teaching, student development, etc.) and (2) talents
education and training in the age of AI. In this book, the review and research of AI in
education place an emphasis on the cross-border integration of AI technologies into
authentic educational practices and core educational services. AI in education will
lead to an improvement in both teachers and learners’ efficiency, provide innovative
educational services, and create a new ecology of education.
1.3 Typical Areas of AI in Education
AI in education is a basic approach to realize future education and a fundamental

goal that future education pursues [24]. However, researchers and educational prac-
titioners should avoid simply implementing and superimposing the components in
traditional education with AI technologies. A more reasonable way to transform
education with AI is to analyze existing problems in educational practices that specific
AI technologies can contribute in addressing them. The two primary services in
education are teaching and management. The completion of teaching relies on the
analysis, design, and implementation of core components of instructional processes.
Educational management is typically achieved through analysis and diagnosis of
problems emerged from the ongoing implementation of teaching.
In traditional educational practices, researchers and teachers pay attention to the
following aspects: the design of learning environments, learner analysis, identifying
learning content, selection of instructional strategies and tools, instructional evalu-
ation and feedback as well as management and decision-making based on the eval-
uation. In the age of AI, learner’s increasingly growing needs make these aspects
even more complicated. For instance, in a traditional teaching context, the teacher
designs and organizes learning content based on his/her subjective understanding of
what his/her students have learned in the previous classes and problems reflected
in students’ homework. However, nowadays, teaching and learning are not confined
to informal learning in the classroom and knowledge from textbooks. Learners can
access personalized learning materials anytime and anywhere with any device, which
will become a fundamental approach to future education [25]. In this context, even at
the same time, in the same place, learners learning the same subject may have entirely
different needs of learning materials. Both teachers and educational administrators
will have to adjust and adapt themselves to the materials and services prepared for
students. From the teachers’ viewpoint, they need to be equipped with the skills to
conduct an individualized analysis of students’ characteristics and capacities. With
the results obtained from this analysis, teachers design appropriate learning envi-
ronments, provide support to students’ learning process, and make evaluation and
diagnosis based on multi-source data. Meanwhile, teachers also need to make changes
1.3 Typical Areas of AI in Education 5
Fig. 1.1 Key aspects of AI in education
to their routines of teaching and research in order to offer more personalized guid-
ance to students. While for educational administrators, they are supposed to make
educational decisions based on the analysis of specific groups and regions.
In the age of AI, the grand challenges we are confronting are how to perceive,
adjust, evaluate, and manage the processes of teaching and learning with AI tech-
nologies. Although these challenges are quite similar to the ones in traditional educa-
tional practices, it is almost impossible for teachers and administrators to solve these
problems based on natural intelligence as they did in the past.
The needs to address complex problems, such as conducting extensive data anal-
ysis and modeling and making big data-driven decisions, match the advantages that
AI technologies go beyond human intelligence. In many fields, there are successful
practices with AI technologies, such as smart home and emotional recognition. It is
promising that with AI technologies, we will be able to tackle major challenges and
issues in education.
With the above needs in concern, Fig. 1.1 presents the following key perspectives
regarding how AI technologies will foster education: (1) intelligent learning environ-
ments; (2) intelligent support to the learning process; (3) intelligent learning assess-
ment; (4) intelligent teacher assistant; and (5) intelligent educational management
and services.
The intelligent learning environment is the foundation of intelligent education,
which is built on pervasive computing technology and perceptual computing tech-
nology. These technologies enable the perceptual function of learning environments
so as to meet diverse needs and offer personalized services in a ubiquitous learning
context, which includes the construction of the smart campus, campus security detec-
tion and early alarming, smart classroom, unsupervised library, and AI-supported
writing system.
Intelligent support to the learning process is the service that AI technologies
provide to learners. AI can offer learners with intelligent support for learner’s char-
acteristics diagnosis, the appropriate learning content, and the right learning tools. It
is composed of the following services: intelligent representation of knowledge and
6 1 Introduction
capacity structures, intelligent diagnosis and recommendation, cognitive load moni-

toring and early alarming, virtual inquiry learning environment, intelligent subject
learning tools, intelligent tutoring learning based on brain science, intelligent robot
partners and toys as well as special education assistants.
Intelligent learning assessment aims to diagnose students’ knowledge progress,
physical situation, mental state, and provide feedback. These services include the
diagnosis and evaluation of learners’ problem-solving ability, the monitoring and
early alarming of learners’ mental health, the monitoring and improvement of
learners’ physical situation, assessment of smart classrooms, learners’ development
planning.
From the teacher’s point of view, intelligent assistant alleviates teacher’s working
load in the era of AI. Primary services that intelligent teaching assistant offers
in this book include intelligent tutoring, intelligent support to answer questions,
intelligent support to compile questions for a quiz, automatic scoring, personalized
support to students’ homework, personalized assessment report generation, teaching
and research report generation, automatic instruction design generation, teaching
summary generation for each semester and school year, and AI-facilitated good
teacher support.
Finally, to provide intelligent management service to educational administrators,
we describe how we can use AI technologies to promote educational equity based on
massive educational data collection and analysis. The services that AI technologies
supply also include intelligent educational decision-making, guidance to improve
regional quality of education, dynamic monitoring of educational quality, customized
educational services, and intelligent health service assistants. These services are
elaborated in Chapters three to seven.
References
1. China Institute of Electronic Technology Standardization (2018) Artificial intelligence stan-

dardization white paper
2. The State Council of China (2017) The next generation artificial intelligence development
planning. http://www.gov.cn/zhengce/content/2017-07/20/content_5211996.htm. Accessed 1
Feb 2018
3. National Science and Technology Council (2016) Preparing for the future of artificial
intelligence. Washington, D.C. https://www.nitrd.gov/pubs/national_ai_rd_strategic_plan.pdf.
Accessed 1 May 2018
4. National Science and Technology Council (2016) The national artificial intelligence research
and development strategic plan. Washington, D.C. https://www.nitrd.gov/PUBS/national_ai_
rd_strategic_plan.pdf. Accessed 1 May 2018
5. Stanford University (2016) Artificial intelligence and life in 2030. https://ai100.stanford.edu/
sites/g/files/sbiybj9861/f/ai_100_report_0831fnl.pdf. Accessed 1 May 2018
6. Luckin R, Holmes W, Griffiths M et al (2016) Intelligence unleashed: an argument for AI in
education. Pearson, London
7. Becker AS, Freeman A, Giesinger HC et al (2016) NMC/CoSN Horizon report (K-12 edition).
Austin, Texas: The New Media Consortium
References 7
8. Freeman A, Becker AS, Cummins M et al (2017) NMC/CoSN Horizon report (K-12 edition).
Austin, Texas: The New Media Consortium
9. Becker AS, Cummins M, Davis A et al (2017) NMC/CoSN Horizon report (Higher education
edition). Austin, Texas: The New Media Consortium
10. Zhang Y (2004) Principles and application of artificial intelligence. Higher Education Press,
Beijing, China
11. Jia J (2009) Educational technology and artificial intelligence. Jilin University Press,
Changchun, China
12. Yi Z (2017) AI powers people. http://news.163.com/shuangchuang/17/1226/15/D6JG703E0
0019AI4.html. Accessed 1 May 2018
13. Yu S, Wang A (2016) The reform trajectory of AI in education. e-Education Res 10:1–9
14. Yu S (2011) How technology will transform education. e-Education Res 7:1–6
15. Dai R (2007) Science of social intelligence. Shanghai Jiao Tong University Press, Shanghai,
China
16. Skinner BF (1958) Teaching machines. Science 28(3330):969–977
17. Jia J (2018) AI powers education and learning. J Dist Educ. https://doi.org/10.15881/j.cnki.
cn33-1304/g4.2018.01.004
18. Katherine N (2016) These graduate students had no idea their teaching assistant was a
robot. https://www.pcworld.com/article/3067749/these-graduate-students-had-no-idea-their-
teaching-assistant-was-a-robot.html. Accessed 18 Sept 2019
19. Pan Y (2018) AI 2.0 and the development of education. Open Educ Res 5:1–5
20. Yu S (2018) The roles of AI-powered teachers in future education. Open Educ Res 07-179(018)
01-016-13
21. Woolf BP, Lane HC, Chaudhri VK et al (2016) AI grand challenges for education. AI Mag
34(4):66–84
22. Zhang K, Zhang J (2017) The new area, misunderstanding, blind area, and restricted zone in
studies and practice of AI in education. Distance Educ J 35(5):54–63
23. Yan Z, Tang X, Qin X et al (2017) The connotations, key technologies, and application trends
of AI in education. Distance Educ J 35(1):26–35
24. Wu Y, Liu B, Ma X (2017) Constructing an ecosystem of AI in education. Distance Educ J
35(5):27–39
25. Chen M, Yu S (2015) The design of recommendation system for learning contexts in ubiquitous
learning environments. e-Education Res 36(4):76–82
Chapter 2
An Overview of AI
2.1 Basic Approaches in AI
In the 1940s to 1950s, a group of scientists from different fields began to explore the
possibility of making an artificial brain. At the 1956 Dartmouth Conference, AI was
officially established as an academic discipline [1]. Over a half-century of develop-
ment, scientists have made various attempts to explore different directions. The field
of AI has thus given birth to many schools. At present, the primary approaches in AI
are behaviorism, symbolism, and connectionism [2].
Behaviorism
Behaviorism believes that AI originates from the perception and control of behaviors
and actions. This approach is built upon the principles of cybernetics. Researchers
who follow this approach simulate the intelligent performance of human behaviors
with an evolutionary method. The perceptual and action model, intensive learning,
brain-inspired computing, and biological intelligence algorithms are exemplary
achievements of this approach.
Behaviorism derives from a school of psychology at the beginning of the twen-
tieth century. The original idea of this approach is that behaviors are a combination
of various physical reactions that organisms use to adapt to environmental changes.
Therefore, behaviors are the subjects of psychological research [3]. In AI, followers
of behaviorism believe that intelligence is the adaptation to the complex environ-
ment of the outside world. This adaptation depends on both perception and actions.
AI constantly regulates their behaviors through interactions with the external envi-
ronment to achieve the goal of sustained adaptation. Similar to human intelligence,
AI evolves gradually through imitating behaviors that are evolvable.
Robot Willow Garage PR2 [4] is a typical case of behavioral AI. Reinforcement of
learning algorithms enables the robot to learn to complete a variety of simple tasks,
such as playing Lego toys, opening bottle caps, assembling toy planes, and plugging
the shoe tree. In these processes, PR2 experiences failures, adapts to changes like

10 2 An Overview of AI
humans, and grasps new skills. Similarly, Spot, a Google robot dog, can walk, run,
complete the movement of getting over all kinds of terrain, and even quickly adjust
its postures in the face of sudden shocks [5]. These two examples embody the char-
acteristics of behaviorism approach of AI, which adapts to different contexts through
perception and action control and evolves through interaction with the environment.
Symbolism
Symbolism, also known as the logic, psychology, or computer school, is an intelli-
gent simulation method based on logical reasoning. Symbolic AI was the dominant
paradigm of AI research from the 1950s to the 1980s. Representative accomplish-
ments include machine proof, expert system, and knowledge engineering. Symbolism
holds that AI originates from mathematical logic. Symbolic AI uses the law of logical
reasoning to imitate human intelligent activities to achieve the simulation of the
human brain’s functions.
Leading researchers of symbolic AI include Allen Newell, Herbert A. Simon, and
Nils J. Nilsson. They believe the primitive of human cognition and thinking is the
symbol. In this way, knowing is a process of symbolic manipulation. Symbolic AI is
characterized as a heuristic symbolic operation process based on previous knowledge.
Followers of symbolic AI think that it is possible to establish the same system with
symbols for machine intelligence.
One of the great achievements of symbolic AI is Deep Blue. The supercom-
puter from IBM, which beat chess champion Garry Kasparov in 1997, represented
a milestone in the history of AI [6]. Later, IBM launched Watson, who successfully
defeated the most dominant player in the American live TV show Danger Edge in
2011. Today, it is dedicated to providing AI solutions for a wide range of fields, such
as in the healthcare industry, to provide patients with the best treatment options.
Connectionism
Connectionism believes that AI originates from bionics. The underlying principle is
to simulate the function of the brain by constructing neural networks and investigating
the connection mechanism and learning algorithms between neural networks. That
is why it is also called bionics or physiology. The school of connectionism starts
their research with the neurons and examines both the neural network model and the
brain model.
The core idea of connectionism is to treat the primitive of intelligent activities as
a neuron. Based on this assumption, the intelligent activity process can be regarded
as the activity process of nerve cells, which is dynamic and evolving. Advocates
of connectionism believe that the structure of neural networks is closely related to
intelligent behaviors. Different structures exhibit different functions and behaviors.
AI emphasizes the simulation of the structure of the human physiological neural
network.
Recently, research on the development and application of deep neural networks
has attracted much attention in AI. The most typical cases are research on convolu-
tional neural network (CNN) and recurrent neural network (RNN). The significant
advantage of CNN lies in its function of feature extraction, which is widely used for
2.1 Basic Approaches in AI 11
two-dimensional image recognition with distorted invariances, such as image recog-

nition and handwriting recognition. In contrast, the hidden layers of RNN have a rela-
tionship with each other as input and output. RNN is usually used in the processing
of sequence data, such as language recognition and machine translation.
One of the most famous examples of deep neural networks is the sensational
AlphaGo, an AI Go program developed by Google DeepMind in London. AlphaGo
combines Monte Carlo tree search and deep neural networks to beat the opponent
by matching and learning the past chess data of professional chess players. In March
2016 and May 2017, AlphaGo defeated world champion—Korean professional chess
player Se-dol Lee (9th-dan player, which is the highest grade in Go) and Chinese
professional chess player Jie Ke (9th-dan player), respectively. Right after that, the
AlphaGo’s team introduced AlphaZero that does not involve any data from the human
[7]. AlphaZero can learn the rules of Go by playing chess with itself, without the data
of human Go players. It took merely 40 days to exceed all previous versions created
by the AlphaGo team. It is worth mentioning that after playing chess with Jie Ke,
Deepmind announced AlphaGo’s retirement because the human chess players were
no longer able to pose a challenge to the AI chess player. The team decided to apply
AlphaGo’s technology to a broader range of areas in the future, such as health care.
Besides, computer vision is also an important application of deep learning and
neural networks. For example, the former head of Google AI China Center, Dr. Fei-
Fei Li, a famous computer vision scientist, has set up a large number of graph database
ImageNet with other scientists [8]. They built a comprehensive and robust dataset for
research in computer vision that made outstanding contributions to the development
in this field. Dr. Enda Wu, who is the director of the Artificial Intelligence Laboratory
at Stanford University and a former Google scientist, has developed an artificial
neural network to learn about how to identify videos about cats by watching YouTube
videos for a week [9]. These products are typical cases of AI with the connectionism
approach.
All in all, these three approaches have promoted the development of research
in AI. In practice, they formulated a unique system of problem-solving methods
and corresponding successful examples of application. Meanwhile, researchers also
realize that each approach has its advantages and disadvantages. At the same time,
the emergence of a large number of AI products and applications has also made the
vision that AI is capable of bringing about practical convenience in people’s real life
into reality.
2.2 Major AI Technologies
AI is not a single technology but a group of technologies that work together as a whole.
This section begins with a general technology framework for the application of AI in
education. Right after that, there is an overview of major AI technologies, including
machine learning, cloud computing, knowledge graph, natural language processing,
computer vision, human–computer interaction, virtual reality and augmented reality,

intelligent control, and robots.
A General Framework of AI in Education

With the continuous development of the research and exploration of AI, more and
more enterprises and research institutions have designed and developed different
systems to support the application of AI in education. We summarize a general frame-
work of AI in education based on these systems, as shown in Fig. 2.1. Overall, there
are four different levels, including the data and platform layer, the AI engine layer,
the deep education application layer, and the application scenario layer. Detailed
information about each layer is elaborated below.
Data is fundamental to the application of AI in education. The educational data
module is responsible for the collection and collation of multivariate data. These data
are further divided into different categories. One category is data based on traditional
Fig. 2.1 General technology framework of AI in education

2.2 Major AI Technologies 13
educational materials, such as textbook learning, tutoring exercises, and instruction

plans. Online learning data is collected from online learning platforms, including
learning process data and online learning assessment data. Internet data is collected
from the encyclopedia and discussion forms. There are also some other education-
related data, such as data generated in the services of educational management as
well as data collected from wearable devices.
The increasing computing power provided by various computing platforms is
a guarantee for AI. A typical computing platform consists of the following parts:
(1) the cloud platform which provides hardware support to the entire platform,
including server configuration, network settings, and storage equipment configu-
ration; (2) the big data analysis platform which analyzes and processes education
big data with frameworks like Hadoop, Spark, and Flink; (3) the machine learning
algorithm framework which supports the deep learning framework, such as Tensor-
Flow, Caffe, Theano, Apache MXNet, and CNTK; and (4) functional components
that support the operation of big data platforms such as security monitoring.
Based on the data from the educational data source module and the computing
power provided by the computing platforms, we can construct various computing
engines to meet the needs of authentic educational practices. Well-constructed algo-
rithms are the backbone of AI. Different AI algorithms can be developed for similar
educational scenarios. At present, dominant algorithms in AI are from machine
learning, including supervised learning, unsupervised learning, semi-supervised
learning, and reinforcement learning algorithms. With the continuous development
of deep learning technologies, all kinds of deep neural network algorithms have been
developed and applied to authentic educational scenarios. Meanwhile, the probability
theory and statistics also provide a variety of probabilistic graph models for educa-
tional problem-solving. Data mining algorithms, such as the association rule mining
algorithms, can assist in discovering intelligence from a data perspective better. Simi-
larly, diagram analysis algorithms contribute to the processing and parsing of graph
data.
A variety of AI engines are built on these algorithms mentioned above. Among
them, the affective computing engine identifies and analyzes the emotional state of
students with different AI algorithms. The knowledge computing engine is respon-
sible for organizing and constructing structured domain knowledge as well as
conducting in-depth mining analysis of the domain knowledge. Based on multi-
variate heterogeneous educational data, the knowledge engine uses knowledge
graph technology to construct knowledge maps in different subjects. The percep-
tual computing engine enables perceptual ability like human beings through natural
language processing, speech recognition, image processing, video analysis, and
other technologies. Compared with the perceptual computing engine, the cognitive
computing engine is to realize the human like comprehension, thinking or making
decision with the help of natural language comprehension, visual image comprehen-
sion, motion behavior comprehension, semantic reasoning, intelligent questioning,
and answering. The intelligent recommendation engine is built with algorithm models
in recommendation system, which supplies students with a personalized learning path
and learning resources.
The ultimate goal of applying AI in education is to improve learning efficiency and

education quality. For example, with the support of AI, we can generate personalized
learning paths and appropriate resources for students. Based on the assessment data
about students and analysis of the relationship among different knowledge with
computing engines, we can offer a more precise diagnosis of the difficulties that they
encounter. Similarly, the automatic generation of learning resources for students
needs to combine the collation of domain knowledge in the knowledge computing
engine and the GAN network generation algorithm in deep learning. The monitoring
and prediction of the mental health of students rely on the emotional computing
engine, which helps to identify and analyze students’ emotions.
Apart from the major parts mentioned above, there are three additional compo-
nents in a typical system of AI in education, which are application access terminal,
target population, and intelligent learning environments. The application access
terminals focus on how to use a variety of devices in different educational contexts,
among which robots have been the one that attracts the attention of both researchers
and practitioners. The AI system offers different services to main participants
involved in the process of teaching and learning, including teachers, parents, students,
administrators, and policymakers. The third one is intelligent learning environments,
such as smart classroom and smart campus, which integrate intelligent technologies.
Cloud Computing and Big Data

The new generation of AI is based on big data. However, there is no uniform definition
of big data so far. According to Gartner [10], big data is high-volume, high-velocity,
and high-variety information assets that demand innovative processing models to
enable more advanced decision-making and process automation. McKinsey Global
Institute defines big data as a collection of data that is large enough to go beyond
the capabilities of traditional database software tools, in terms of its acquisition,
storage, management, and analysis ability [11]. Wikipedia defines big data as a large
or complex dataset that is not sufficient for traditional data processing applications to
process [12]. Even though the term is loosely defined, we can summarize some shared
features from these statements, which are massive data size (volume), rapid data flow
(velocity), diverse data types (variety), and low-value density (value). Moreover,
veracity is added as the fifth feature of big data [13], as shown in Fig. 2.2.
Computing and storage are two significant challenges that big data is facing
now and seeking to address in the future. Cloud computing is built on parallel
computing, which refers to the process of using multiple computing resources to
solve computational problems simultaneously. Compared with general computing,
parallel computing can execute multiple instructions concurrently to address complex
problems with a higher calculation speed and a more extensive data scale. Cloud
computing arises after parallel computing. Currently, National Institute of Stan-
dards and Technology (NIST) defines it as a model for convenient and on-demand
network access to a pool of configurable resources that can be provided with minimal
management effort is widely accepted by researchers [14].
As one of the latest hot topics in computer science research, cloud computing is
highly reliable, versatile, and expansible. The unlimited storage size that the cloud
Fig. 2.2 5 V features of big

data
offers enables unprecedented computing power and storage. Users can use a portion
of the cloud services according to their needs and pay for those similar services.
Users do not need to worry about the extra fees involved in other relevant cloud
services. Also, the virtualization of cloud computing empowers vast data computing
and storage without the restriction of geographic location.
Cloud computing and big data complement each other and promote the develop-
ment of each other to drive the development of AI technologies. Researchers point
out that with big data and cloud computing technologies, we can discover students’
implicit needs and improve the service of online learning based on the collection,
identification, and analysis of learning data [15]. Rong Zhang believes that big data
and cloud computing technologies can assist in obtaining regional education data
and provide vital support for scientific decision-making [16]. Ling Jin argues that
the innovative and effective use of big data can promote the advancement of the inte-
gration of information technology in teaching and learning [17]. With the assistance
of big data and cloud computing technologies, we can develop new solutions to many
long-standing issues in education.
Machine Learning
Machine learning involves computer science, statistics, neural networks, proba-
bility theory, optimization theory, complexity theory, and many other areas. Machine
learning is one of the essential methods in AI research. The core idea of machine
learning is to design and analyze algorithms that support machines to learn “auto-
matically” to continuously improve the performance of the algorithms through more
data and previous learning experience. There are three different methods in machine
learning, which are supervised learning, unsupervised learning, semi-supervise
learning, and reinforcement learning (as shown in Fig. 2.3).
Fig. 2.3 Technology framework of machine learning
Supervised Learning: Supervised learning refers to the use of a given labeled

dataset to train the machine learning model and the further use of the model to predict
new data. Supervised learning algorithms can be grouped into two broad categories:
classification and regression.
Unsupervised Learning: Different from supervised learning, the training dataset
is not predefined in unsupervised learning. More specifically, the data is self-
consolidated and summed up to construct a model depicting underlying rules and
features in unlabeled data. Clustering is one of the typical unsupervised learning
algorithms.
Semi-supervised Learning: Semi-supervised learning is a method between
supervised learning and unsupervised learning. In the semi-supervised learning
dataset, some data are labeled. However, a majority of the rest are unlabeled. Data
in the semi-supervised dataset is not completely random. Local features of the data
are excavated with labeled data. The overall distribution features of the dataset are
depicted with unlabeled data. With both the labeled and unlabeled data, a better
model can be obtained to predict new data.
Reinforcement Learning: It is a method that takes environmental factors into
account in the training process, which continually adjusts the model and provides
feedback based on stimuli in external environments. With the guidance of the current
model, it chooses the next step that matches the needs to refining the model best and
keeps adapting the model until it converges finally.
Besides, machine learning includes a variety of other relevant terms, such as deep
learning, transfer learning, and evolutionary learning. These different terms also
reflect the diversity and comprehensiveness of machine learning algorithms.
The essence of machine learning is to use the computer to discover rules from raw
data, identify patterns of data, and use them for prediction [18]. Machine learning is
widely used in many aspects of teaching and learning. Concerning student modeling,
Eagle and colleagues integrated student parameters in the tradition Bayesian knowl-
edge tracking model in the intelligent tutoring system to predict the weight of indi-
vidual’s difference and performance based on their activity data (including the data of
reading performance and conceptual knowledge preparation test). Findings suggest
that the data of text reading is of great use to predict students’ learning and perfor-
mance in the intelligent tutoring system [19]. Besides, researchers also used multi-
strategy machine learning, such as inductive reasoning and analogical reasoning, to
construct advanced learning models. They used these learning models to learn and
discover the attributes of inconsistent behaviors in the learning process. Based on
these attributes, the tutoring system monitors student’s behaviors to prevent students
from re-behaving inconsistently through appropriate methods (such as strengthening
teaching and practice) [20].
Regarding predicting students’ learning performance, Hachey and others used the
two-element logistic regression algorithm to analyze 962 students’ previous online
learning experience and their GPAs. The results showed that students’ previous
learning experience could serve as an indicator of the performance of their online
learning [21]. In terms of early warning of the dropout, Aulck and colleagues used
machine learning to keep track of the performance data and demographic data of
32,500 students, including the methods of regularized logistic regression, K-nearest
neighbor algorithm, and random forest algorithm. With these methods, they tested
the best elements for predicting dropout. They also compiled a series of indicators
of dropout, including mathematics, English, chemistry, and psychology courses, as
well as the school enrollment and birth date [22]. Moreover, Aher and Lobo used
data mining techniques (such as clustering and association rule algorithms) in course
recommendation systems to recommend courses for students based on courses they
had taken in Moodle and their interest [23]. Overall, machine learning will assist in
teaching and learning by providing more intelligent and human-centered services.
Natural Language Processing

Natural language processing (NLP) is a branch of AI and linguistics, which primarily
investigates various theories and methods of effective communication between
machines and human beings with natural language [24]. The major categories of
NLP include semantic analysis, lexical labeling, syntactic analysis, text classifica-
tion, information retrieval, information extraction, machine translation, sentiment
analysis, and so on.
The process of NLP involves two steps, namely natural language comprehen-
sion and natural language generation. Natural language understanding refers to the
transformation of natural language into a form that computers can understand and
manipulate. Natural language generation refers to the transformation of programs
and data in computers into natural languages, including text planning, statement
planning, and implementation. The NLP technology framework is shown in Fig. 2.4.
The ultimate goal of NLP is to enable human communication with computers in
natural language.
Fig. 2.4 Technology framework of NLP
At present, there are many applications of NLP in education. For instance, the
intelligent language teaching system (such as the LAIX new media company) can
identify user’s pronunciation and return a real-time score as feedback to help learners
improve their pronunciation through natural language processing and other technolo-
gies. For automated scoring of answers to subjective questions, Pigai generates live
scores, comments, and content analysis results of student’s essays. He indicated that
Pigai could not only motivate students to improve their writing but could also alle-
viate teacher’s workload in English essay scoring [25]. Besides, Luo and others used
NLP and sentiment analysis technologies to dig into students’ evaluation of their
teacher’s teaching and generate a summary for the teacher as a reference to improve
his or her teaching [26].
However, there are still many challenges in NLP. For example, the boundary
of words in a sentence needs to be defined more precisely. Furthermore, the exact
meaning of polysemy and ambiguous words in a specific context needs to be clarified.
Also, the fuzziness of syntax needs to be eliminated using the context. Defective or
irregular inputs (such as non-standard pronunciation in input and blurred speech)
need to be dealt with as well. Finally, the potential meaning of language needs to be
understood with specific behaviors.
Computer Vision
Computer vision, also known as machine vision, uses computers to analyze and
understand the world. In machine learning, computer vision, NLP, and speech recog-
nition are recognized as the three hot directions. As a subject that has achieved
considerable progress in AI, computer vision itself includes many different research
directions, of which the most popular are object recognition and detection, motion
and tracking, scene reconstruction, and image recovery [27].
With the development and application of deep neural networks, the advantages of
convolution neural networks in image feature extraction gradually give prominence.
Compared with the traditional model, it involves fewer connections and training
parameters and shows stronger adaptability. Therefore, computer vision has a wide
range of applications in handwriting recognition and image recognition. With respect
to driverless technology, the driverless car is one of the foci of Internet compa-
nies and traditional car companies in AI research. Driverless cars rely primarily
on computer-based intelligent drivers in their cars, which is inextricably linked to
computer vision. Specifically, the core of the driverless technology is a computer in
the car which generates real-time pictures and videos for itself to recognize and learn
to drive. Besides, pedestrian detection, road identification, and pattern recognition
in the driving process are also inseparable from computer vision technology [28].
Computer vision has also added vitality into innovative educational teaching
methods. Camera-based interaction is the main application of a computer vision
algorithm, which enables human–machine interaction by analyzing the user’s action
information captured by the camera. For example, Delegado-Mata have categorized
children motor skills of different ages, analyzed the infrared image from Wii (the
home console from Nintendo), applied motion recognition techniques, and then
designed an interactive game to improve children’s motor skills and spatial cognitive
skills [29]. Chen and her colleagues integrated computer vision, behavior recog-
nition, and other technologies to create an interactive intelligent learning environ-
ment for autistic children. In the learning process, the system adjusts the prepared
learning activities based on the observation of children’s behaviors, cognition, and
psychological status to improve their social skills [30].
Human–Computer Interaction
Human–computer interaction is the product of combining computer science
and cognitive psychology. There are some other disciplines involved, including
ergonomics, sociology, physiology, medicine, linguistics, and philosophy. Human–
computer interaction seeks to understand the interactions between the machine
and the user. Human–computer interaction investigates both the existing modes of
interaction between humans and machines and other new models of interactions.
The traditional model of human–computer interaction depends on various external
equipment, such as a mouse, keyboard, capacitive pen, control lever, as well as some
output equipment like printers, monitors, and stereo equipment. With the develop-
ment of technology, such as speech interaction, affective interaction, embodied and
tangible interaction, and brain–machine interaction, novel and diverse interaction
patterns have emerged. More senses are involved in the process, which brings about
more harmony and efficient human–computer interactions. On the one hand, multi-
touch screen technology has promoted the transformation of critical devices to touch
devices. Gesture interaction has gradually become the mainstream. On the other hand,
the progress of speech recognition technology makes language interaction the best
alternative to gesture interaction. With the maturity of speech recognition and natural
language processing technology, such as smart speakers and smart home appliances,
a variety of voice interaction devices and products are now gradually applied in our
daily life due to the natural way of interaction, high input rate, productive infor-
mation transmission, and other advantages. Also, the development of sensors tech-
nology extends gesture interactions to the whole-body postures. The development of
wearable sensor devices enables more compatible interaction models with our daily
life. For example, context-aware technology is now used on a security bracelet to
detect and send distress messages automatically to family members in dangerous
situations, when the user falls. When combined with GPS positioning technology of
mobile phones, it can automatically mute our phones in meeting rooms and cinemas
and shield incoming calls when the user is driving. The perceptual techniques of
posture and gestures are also commonly used in some VR/AR games to optimize the
sensory experience of players and enhance the fun of the game (see HoloLens from
Microsoft for an example [31]).
At present, a variety of human–machine interaction modes are gradually intro-
duced into education. For example, the language interaction mode is widely used in
learner-directed oral English learning, with which learners can practice in the intelli-
gent system anytime and anywhere. The system provides learners with automatic and
timely evaluation and feedback to enhance learners’ oral English. The embodied and
tangible interaction mode is also widely used in the instruction of different subjects,
including the operation of three-dimensional graphics through body sensing equip-
ment, the control of experimental equipment through sensing equipment drawing in
physics, and the manipulation of time corridor through the body sense equipment
[32].
Robot and Intelligent Control

Intelligent control refers to a class of control techniques that use different AI
computing approaches to achieve a goal without direct intervention of humans. The
theoretical foundation of intelligent control involves AI, cybernetics, operational
research, and information theory. Intelligent control measures the controlled amount
automatically and figures out the deviation from the expectation. It collects the infor-
mation from the input and gets the output control accordingly to reduce or eliminate
the deviation as far as possible [33].
In AI research, reinforcement learning is an effective means for intelligent control.
By simulating the learning process of humans, it observes the interaction between
the agent and the external environment and gets trained in the process. Because of the
unpredictable nature of the real environment, it is a complex task for the agent to learn
rules by modeling the space status. Reinforcement learning is not concerned with
the specific process and details of problem-solving. The task designer will provide
feedback according to the objective function, measure the performance of each step
in the interaction process, and discover the best behaviors through trial and error [34].
Therefore, it is ideal for solving complex problems without a transparent process.
For instance, Boston Dynamics developed a four-legged walking robot, developed
by Boston Dynamics, which can perform tasks at the office, at home and outdoors,
such as grasping objects and climbing stairs [35].
The robot, as one of the typical carriers of intelligent agents, plays an essential
role in fostering different applications in learning. It combines a variety of prac-
tical scenarios with AI through intelligent control technology, replacing humans
with repetitive, complex, and dangerous work. At the same time, as a carrier of AI
technologies, robots can also integrate various artificial intelligence technologies in
different fields. In the field of education, there is an educational robot developed by

the Advanced Innovation Center for the Future Education of Beijing Normal Univer-
sity, which integrates the basic concepts of different disciplines into the robot system
through the knowledge graph technology [36]. The robot communicates and interacts
with users through natural language processing technology. Two primary functions
of the robots are knowledge delivery and emotional companionship. To be specific,
the robot can monitor students’ progress of knowledge advancement in real time and
make corresponding learning resources recommendations for weak areas. Besides,
the robot can also monitor students’ emotional experiences and provide real-time
emotional care to encourage students to overcome difficulties. Educational robots
are also used to help children with special needs and promote their healthy growth
[37].
Knowledge Graph
In 2012, Google launched its first version of the knowledge graph, setting off a wave
in both academia and industry. Nowadays, the knowledge graph has become one of
the research hot spots.
A knowledge graph is essentially a graph-based data structure consisting of nodes
and edges [38]. Nodes represent entities in real world; the lines between nodes
represent the relationships between entities; different entities are connected through
the entity-relationship-entity group mode to form a networked knowledge structure
[39] so that information resources are easy to use. It also provides a way to analyze
education issues from the perspective of the relationship.
As the knowledge graph technology matures, it can be used in education to
support teaching and learning more widely. The knowledge graph reveals the essen-
tial attribute behind learning content from the perspective of the relationship, which
is beneficial to students’ mastering of basic concepts and the relationship among
them. With the domain knowledge graph, the system can provide automatic question
and answer service according to students’ needs. For example, when the student
asks, “Where does photosynthesis occur,” the system can respond quickly to offer
the student with correct feedback. Besides, with the help of the knowledge graph,
the system can make intelligent recommendations of relevant learning resources to
students. When the student searches for resources about “the application of congruent
triangles,” the algorithm will infer the real intention behind the student’s manipulation
and recommend relevant resources about congruent triangles to the student.
Virtual Reality and Augmented Reality

Virtual reality (VR) and augmented reality (AR) is the use of images, video, 3D
models, and other technologies to create a digital virtual world with integrated senses
[40]. VR/AR integrates a variety of intelligent technologies, including computer
graphics, simulation technology, multimedia technology, AI technology, computer
network technology, parallel processing technology, and multi-sensor technology.
The difference between the two is that VR technology focuses on the creation of a
virtual environment, while AR technology focuses on the superposition of a virtual
environment and real environment. Virtual reality and augmented reality technology
have rich applications and significance. VR and AR technologies enable real, secure,
and targeted simulation solutions for scenarios that is limited in resources, requires
expensive equipment, or is risky or hard to demonstrate. In education, AR/VR tech-
nologies also play a vital role in supporting teaching and learning, professional
training, and virtual campus construction.
First of all, for scientific research, the use of VR/AR technology saves the cost of
expensive equipment and avoids experimental risks by breaking the space and envi-
ronmental constraints. Second, for the professional skill training, VR/AR technolo-
gies take advantage of the immersive and interactive environment to offer students
with real and suitable experience in learning. Finally, in the virtual campus, virtual
reality and augmented reality can provide different levels of experience. Low-level
service refers to virtual visit of the campus landscape, facilities, school affairs, and
campus life. The high-level service refers to interactive distance education services.
By providing a mobile learning space for schools to expand their learning options,
it also supplies the community with training opportunities to create economic and
social benefits.
At present, with the increasing demand for VR/AR products, emergent chal-
lenges that next-generation AR/VR technologies need to address are the portability
of equipment, the freedom of interaction, and the dizziness control of the immersive
experience.
2.3 Typical Open-Source AI Projects and Cloud Services
Open-source AI solutions make it easier and faster for developers to get familiar
with and apply various machine learning and artificial intelligence algorithms as
well as to personalize them according to needs. Cloud services, on the other hand,
provide services for a range of functions, such as computing and storage. Users
merely need to pay for the functions and features they need, which reduces the cost
of development at the bottom. Besides, many open-source knowledge databases also
provide researchers and developers with abundant data on domain knowledge graphs.
Below are a few typical AI open-source projects and cloud services.
Baidu PaddlePaddle
As an easy-to-learn and easy-to-use distributed deep learning platform,
PaddlePaddle, launched by Baidu, has provided deep learning algorithm support
for many products [41]. For example, when it comes to computer vision, the CNN
module automatically identifies and classifies the subjects in the image. In terms
of natural language understanding, the LSTM interface provided by PaddlePaddle
can analyze the positive emotion and negative emotion of commentators from the
comments about the film. At the same time, we can use various functions in the
platform to analyze the features of users and different movies, review scores, and
predict new users’ evaluation of different movies to build a film recommendation
system.
2.3 Typical Open-Source AI Projects and Cloud Services 23
In addition to abundant algorithm services, PaddlePaddle is featured by its high

usability. Whether the user has learned deep learning algorithms or not, the platform
enables users to make calls to various deep learning algorithms through intuitive,
flexible data interfaces, and model configuration interfaces. Also, it is easy for instal-
lation. Taking the PIP installation as an example, users only need to input “Pip install
PaddlePaddle” and wait for the installation to be used locally.
Tencent AI Open Platform

Tencent brings together top technologies, professional talents, and industry resources
to provide a lot of AI services, including OCR, face recognition, image recognition,
image special effects, semantic analysis, emotional analysis, machine translation,
and speech recognition and synthesis [42]. So far, Tencent AI offers many industry
solutions, such as smart recruitment and intelligent hardware.
Tencent’s AI open platform provides developers with a development kit for PHP
and Python, which allows them to get the SDK and start calling quickly. For instance,
developers can refer to test_nlp_texttrans.py for configuration and use after installing
Python and the SDK source package. At the same time, Tencent AI open platform
welcomes developers to contribute SDKs in various languages.
iFLYTEK Open Platform

As the world’s first open, intelligent interactive technology service platform,
IFLYTEK Open Platform [43] is a one-stop intelligent human–computer interaction
solution for developers. The platform integrates the technical achievements of speech
recognition, speech synthesis, and other technologies from iFLYTEK Research Insti-
tute at University of Science and Technology of China, the Xunfei Voice Lab at China
University of Science and Technology, and the Tsinghua University. At the same
time, it introduces the most advanced AI technology in China and abroad (such as
face recognition) and cooperate with academics and industry to create a new mobile
ecosystem with the voice as the core.
At present, the open platform provides developers with many services, such as
voice synthesis, voice recognition, voice wake-up, semantic understanding, face
recognition, personalized ringtone, and mobile application analysis. Users can use
the SDK of Android, iOS, WP8, Java, Flash, Windows, Linux, and other platforms
through a variety of access methods. Users do not need to worry about the oper-
ational difficulty and complexity. Currently, iFLYTEK’s open platform has been
applied to smart TV, wearable devices, smart car as well as applications such as
Didi taxi, Amap, and 58.com, allowing users to experience the world’s leading voice
technology directly and integrate it into products rapidly.
Alibaba Cloud
Founded in 2009, Alibaba Cloud (Aliyun) has provided services to enterprises, devel-
opers, and government agencies in over 200 countries and regions. It has opened
45 available areas in 18 regions around the world, providing reliable computing
support to billions of users worldwide [44]. The solutions provided by Alibaba Cloud
cover cloud computing, big data, security, and other fields. The major functions it
offers include elastic computing, cloud storage, data transmission, database disaster
recovery, government cloud security, Internet financial security, big data warehouse,
and data integration on cloud.
When users use Aliyun, they pay by volume, the step billing mode. The more a
customer uses, the lower the cost. This billing encourages large volume and attracts
loyal users. In addition, Alibaba has also developed the ET brain to solve problems
that are difficult for humans to make decisions [45]. At present, it has acquired
intelligent voice interaction, image/video recognition, machine learning, sentiment
analysis, and other skills.
Currently, Aliyun has offered a variety of services. For example, the open-source
project Apache RocketMQ is based on Java distributed message and streaming
computing platform. It provides highly reliable news publishing and subscription
services with low latency. Apache RocketMQ has now been applied in finan-
cial payments, e-commerce, express logistics, advertising and marketing, instant
messaging, and many other fields.
Amazon Web Services

Amazon Web Services (AWS) is a secure cloud service platform that provides
computing power, database storage, content delivery, and other capabilities to help
the business expand and grow [46]. AWS has a wide range of services. Except for
the services mentioned above including computing, storage, database, migration,
networking and content distribution, it also provides a number of other services such
as tools for developers, management tools, media services, machine learning, anal-
ysis, security, identity and compliance, mobile services, AR and VR, application
integration, customer engagement, enterprise productivity, desktop and application
streaming, Internet of Things (IoT), game development, and software development.
As one of the first cloud service providers, AWS uses a pay-per-use pricing model.
AWS pricing is similar to the way that utilities are paid. Users are merely charged
for the services they use. They are not required to pay an additional fee or termina-
tion fee after the service terminates [47]. AWS now has more than 1 million active
users every month in various industries. It is capable of meeting the specific needs of
users in terms of size, operations, security, and compliance [48]. Specifically, AWS
offers proven solutions in a variety of industries, including financial services, digital
marketing, gaming, media and entertainment, government, and educational institu-
tions. It can also provide a wide range of application solutions for Web sites, backup
and restore, development and operations, IoT, blockchain, and more.
Google TensorFlow
TensorFlow, developed by Google, is a deep learning framework that supports a
variety of operating systems such as Linux, Mac-OS, Windows, Android, and iOS.
Users can easily use it to design neural network structures without implementing
C++ or CUDA code in pursuit of efficiency. TensorFlow supports multiple languages
through Simplified Wrapper and Interface Generator (SWIG), which can be devel-
oped using C++, R, and Python. In TensorFlow, the calculation can be represented
as a computation graph, in which each operation will act as a node by edges. Each
2.3 Typical Open-Source AI Projects and Cloud Services 25
node of the computation graph can have any number of inputs and outputs. More-
over, the node can be regarded as an instantiation of the operation. The data flowing
in the edges of the computation graph is called tensor, which is also the origin of
TensorFlow.
As one of the most widely used machine learning frameworks, TensorFlow is
particularly widely used in cloud-based applications, such as Amazon network
services, Microsoft Azure, and Google Cloud Platforms that support TensorFlow-
compatible services. According to its official Web site [49], companies that use
TensorFlow include Airbnb, NVIDIA, Uber, SAP, Dropbox, eBay, Google, Intel,
Coca-Cola, Twitter, and more.
Microsoft Azure
Microsoft Azure [50] is an online public service platform launched by Microsoft,
offering more than 30 kinds of services, including computing, applications, storage,
and analysis. Its features are designed for a wide range of services such as the cloud,
IoT, and big data. Azure has many AI-enabled interfaces. For example, Emotion API
[51], which can automatically recognize faces and analyze expressions and emotions
from pictures and videos [52]. HDInsight has powerful data parallelism and huge
storage space on which developers can cluster Hadoop, HBase, or Apache Storm to
query processed data. Besides, Azure enables users to transform and aggregate data
for interactive analysis. For example, by using workflows for HDInsight jobs, we can
convert data from all shapes and sizes and load results into SQL DB for interactive
analysis and visualization using Power BI [53].
WordNet
WordNet [54] is an English dictionary built and maintained by Princeton University’s
cognitive science laboratory under the direction of psychology professor George A.
Miller. WordNet has received more than $3 million in funding (mainly from govern-
ment agencies interested in machine translation) since 1985 [55]. It was developed
for two purposes: to serve as both a dictionary to provide greater ease of use and to
support automatic text analysis and AI applications. Because of semantic informa-
tion, it is different from a dictionary in daily use. WordNet groups entries according
to their meaning. Each group of notes with the same meaning is called a synset (set
of synonyms). WordNet provides a short definition for each synset and records the
semantic relationship between different synsets.
In WordNet, nouns, verbs, adjectives, and adverbs are organized into a synonym
network. Each synonym collection represents a basic semantic concept. Various
relationships connect these sets. WordNet’s noun network is developed first. The
backbone of the noun network is the level of the relationship, which occupies nearly
80% of the relationship. The top layer in the hierarchy is the fundamental category
of the starting point. In particular, there are 11 abstract concepts, such as entity and
psychological feature. The deepest level in the noun hierarchy also has 16 nodes
[56].
WordNet’s database and the corresponding software tools are issued in accordance
with the BSD license. Everyone can download for local use or query and use it online
for free.
DBpedia
DBpedia [57] is an open-source knowledge base project that extracts structured
content from Wikipedia and exposes the structured information to people on the
Internet [58]. It allows users to query the relationship and nature of Wikipedia-related
resources and even to query content or links from Wikipedia. Most of the entries in
Wikipedia are text that does not have a fixed format. Some information (such as the
table of content, classification, images, geographic coordinates, and links to external
webpages) is structured. Furthermore, the structured information can be extracted in
the DBpedia and placed in a unified dataset for convenient user access.
In September 2013, the DBpedia (version 3.9) data group described 4 million
test questions, of which 3.22 million belonged to a coherent ontology subcategory,
containing 832,000 people, 639,000 landscapes, 116,000 music albums, 78,000
videos, 18,500 video games, 209,000 organizations, 226,000 species, and 5600
diseases. Meanwhile, the DBpedia dataset provides featured tags and data digests in
up to 119 different languages. In total, there are now 24.6 million image links and
27.6 million data linked to external pages, 45 million links to other RDF datasets, 67
million links to Wikipedia classification pages, and 41.2 million YAGO2 classified
data [59].
DBpedia was once commented by the founder of the World Wide Web as one
of the world’s most popular topics for dicentric linked data [60]. The project was
initially opened by people from the Free University of Berlin and the University of
Leipzig, in collaboration with the Open Linked Software Alliance. The first publicly
available dataset was released in 2007. It is distributed through a free license that
allows others to use these datasets freely.
References
1. Nilsson N (2010) The quest for artificial intelligence. Cambridge University Press, New York
2. Cai Z, Xu G (1996) Artificial intelligence and its applications. Tsinghua University Press,
Beijing
3. Xu X, Me J (2004) A review of research on behaviorism artificial intelligence. Control Decis
19(3):241–246
4. Garage W (2019) PR2: Robot for research and innovation. http://www.willowgarage.com/
pages/pr2/overview. Accessed 18 Sept 2019
5. Boston Dynamic (2019) Spot. https://www.bostondynamics.com/spot. Accessed 18 Sept 2019
6. People (2019) Deep Blue bid farewell to chess. http://www.people.com.cn/GB/historic/0923/
4817.html. Accessed 18 Sept 2019
7. DeepMind (2019) AlphaGo Zero: Learning from scratch. https://deepmind.com/blog/alphago-
zero-learning-scratch. Accessed 18 Sept 2019
8. Russakovsky O, Deng J, Su H et al (2015) Imagenet large scale visual recognition challenge.
Int J Comput Vision 115(3):211–252
References 27
9. Wired (2013) The man behind the Google Brain: Andrew NG and the quest for the new AI.
https://www.wired.com/2013/05/neuro-artificial-intelligence. Accessed 18 Sept 2019
10. Gartner (2019) Big data. https://www.gartner.com/it-glossary/big-data. Accessed 18 Sept 2019
11. McKinsey Global Institute (2019) Big data: the next frontier for innovation, competi-
tion, and productivity. https://www.mckinsey.com/business-functions/digital-mckinsey/our-
insights/big-data-the-next-frontier-for-innovation. Accessed 18 Sept 2019
12. Wikipedia (2019) Big data. https://en.wikipedia.org/wiki/Big_data. Accessed 18 Sept 2019
13. China Electronics Standardization Institute (2019) Big data standardization white paper. http://
www.cesi.cn/201612/1692.html. Accessed 18 Sept 2019
14. National Institute of Standards and Technology (2009) NIST definition of cloud computing.
https://www.nist.gov/publications/nist-definition-cloud-computing. Accessed 18 Sept 2019
15. Xing Q, Jiao J, Du Z (2014) Research on interactive application of online education based on
cloud computing and big data. Modern Educ Tech 24(4):88–95
16. Zhang R (2016) The promotion of cloud computing technology in the Age of big data to
regional education development. Adult Educ China 20:122–125
17. Jing L (2013) Big data and teaching reform with information technology. China Educ Tech
10:8–13
18. Yu M, Feng X, Zhu Z (2017) Educational application and innovation exploration of machine
learning from the perspective of artificial intelligence. Distance Educ J 35(3):11–21
19. Eagle M, Corbett A, Stamper J et al (2016) Estimating individual differences for student
modeling in intelligent tutors from reading and pretest data. In: Micarelli A, Stamper J,
Panourgia K (eds) Proceedings of the 13th International conference on intelligent tutoring
systems, vol 9684. Springer, Cham, pp 133–143
20. Huang M, Chiang H, Wu P et al (2013) A multi-strategy machine learning student modeling
for intelligent tutoring systems based on Blackboard approach. Libr Hi Tech 31(2):274–293
21. Hachey AC, Wladis CW, Conway KM (2014) Do prior online course outcomes provide more
information than GPA alone in predicting subsequent online course grades and retention? An
observational study at an urban community college. Comput Educ 72:59–67
22. Aulck L, Velagapudi N, Blumenstock J et al (2016) Predicting student dropout in higher
education. arXiv preprint arXiv:1606.06364
23. Aher SB, Lobo L (2013) Combination of machine learning algorithms for recommendation of
courses in E-learning system based on historical data. Knowl-Based Syst 51:1–14
24. Wang C, Zhang M, Ma S (2007) Application of natural language processing in information
retrieval. Chin J Inf Tech 2:35–45
25. He X (2013) Research on the reliability and validity of English composition grading. Modern
Educ Tech 23(5):64–67
26. Luo Y, Pan Q, Liu L, Zhang L (2018) Design and application of a student evaluation system
based emotion mining. China Educ Tech 4:91–95
27. Tech Briefs (2016) Cambridge researchers take computer vision for test drive. https://www.tec
hbriefs.com/component/content/article/tb/features/application-briefs/24301. Accessed 18 Sept
2019
28. Zaoshu Tech (2017) Computer vision. https://www.zhihu.com/question/19695972. Accessed
18 Sept 2019
29. Delgado-Mata C, Ruvalcaba-Manzano R, Quezada-Patino O et al (2009) Low cost video game
technology to measure and improve motor skills in children. In: AFRICON2009. IEEE, pp 1–6
30. Chen L, Wang G, Zhang K (2017) Design and implementation of human-computer interaction
learning activities to improve social interaction skills of autistic children. China Educ Tech
5:106–111
31. Microsoft (2019) Holoens. https://www.microsoft.com/en-us/hololens. Accessed 18 Sept 2019
32. Li Q, Wang Q (2015) Review on the application of somatosensory interaction technology in
education. Distance Educ J 33(01):48–56
33. Nilsson N (1993) An introduction to intelligent and autonomous control. Kluwer Academic
Publishers, Berlin
34. Kober J, Bagnell J, Peters J (2013) Reinforcement learning in robotics: a survey. Int J Robot
Res 32(11):1238–1274
35. Boston Dynamics (2019) Changing your idea of what robots can do. https://www.bostondyn
amics.com. Accessed 18 Sept 2019
36. Lu Y, Chen C, Chen P, Chen X, Zhuang Z (2018) Smart learning partner: an interactive robot
for education. International conference on artificial intelligence in education. Springer, Cham,
pp 447–451
37. Johnson L, Adams S, Cummins M et al (2016) NMC horizon report: 2016 higher education
edition. The New Media Consortium, Austin, p 46
38. Qi G, Gao H, Wu T (2017) Progress in research on the knowledge graph. Intell Eng 1:4–25
39. Xu Z, Sheng Y, He L et al (2016) An overview of the knowledge graph technology. J Univ
Electron Sci Technol 45(4):589–606
40. Zhou Z, Zhou Y, Xiao J (2015) An overview of virtual reality enhancement technology. Sci
China: Inf Sci 45(02):157–180
41. Baidu (2019) PaddlePaddle. http://paddlepaddle.org. Accessed 18 Sept 2019
42. Tencent (2019) Tencent AI open platform. https://ai.qq.com. Accessed 18 Sept 2019
43. iFlytek Co. (2019) iFlyteck open platform. http://www.xfyun.cn. Accessed 18 Sept 2019
44. Alibaba (2019) Alibaba cloud. https://www.aliyun.com. Accessed 18 Sept 2019
45. Alibaba Cloud (2019) Hello, ET brain. https://et.aliyun.com/index. Accessed 18 Sept 2019
46. Amazon Web Services (2019) Cloud computing and AWS. https://aws.amazon.com/cn/what-
is-aws/?sc_icampaign=aware_what_is_aws&sc_icontent=awssm-evergreen-default-hero&
sc_iplace=hero&trk=ha_awssm-evergreen-default-hero&sc_ichannel=ha. Accessed 18 Sept
2019
47. Amazon Web Services (2019) AWS pricing. https://aws.amazon.com/cn/pricing/?nc2=
h_ql_pr&awsm=ql-3. Accessed 18 Sept 2019
48. Amazon Web Services (2019) Cloud solutions. https://aws.amazon.com/cn/solutions/?nc2=
h_ql_s&awsm=ql-2&solutions-all.sort-by=item.additionalFields.sortDate&solutions-all.sort-
order=desc. Accessed 18 Sept 2019
49. TensorFlow (2019). TensorFlow: an open source software library for machine intelligence.
https://www.tensorflow.org
50. Microsoft Azure (2019) Cloud computing: a reliable intelligent service platform. https://www.
azure.cn/zh-cn. Accessed 18 Sept 2019
51. Microsoft Azure (2019) Cognitive services. https://azure.microsoft.com/en-us/services/cognit
ive-services/emotion. Accessed 18 Sept 2019
52. Microsoft Azure (2016, March 17). Microsoft—the only vendor named a leader in Gartner
Magic Quadrants for IaaS, Application PaaS, and Cloud Storage [EB/OL]. https://archive.is/
PQJWp#selection-5459.0-5459.115. Accessed 18 Sept 2020
53. Microsoft Azure (2019) Solutions for data analysis. https://www.azure.cn/zh-cn/solutions/data-
analysis. Accessed 18 Sept 2019
54. Wordnet (2019) Wordnet: A lexical database for English. https://wordnet.princeton.edu.
Accessed 18 Sept 2019
55. Research Center for Chinese Linguistics at Peking University (2019) The development
of Wordnet. http://ccl.pku.edu.cn/973_sem_spec/WordNet/C-wordnet/wordnet-c-index.htm.
Accessed 18 Sept 2019
56. Wordnet (2019) Wordnet. https://zh.wikipedia.org/wiki/WordNet. Accessed 18 Sept 2019
57. DBpedia (2019) DBpedia. https://wiki.dbpedia.org/about. Accessed 18 Sept 2019
58. Bizer C, Lehmann J, Kobilarov G et al (2009) DBpedia: A crystallization point for the web of
data. Web Semant 7(3):154–165
59. Dbpedia (2019) DBpedia 3.9 released. https://zh.wikipedia.org/wiki/DBpedia. Accessed 18
Sept 2019
60. Berners-Lee T (2008) Tim Berners-Lee talks with Talis about the semantic web. http://talis-
podcasts.s3.amazonaws.com/twt20080207_TimBL.html. Accessed 18 Sept 2019
Chapter 3
Intelligent Learning Environments
Currently, intelligent terminals are gradually becoming popular. Smart terminals,

networks, and platforms are integrated into all aspects of the educational environment
to build a ubiquitous computing space. With powerful cloud computing technologies,
the boundary between virtual space and real space is becoming blurred. The educa-
tional environment is an intelligent ternary integration where the physical space,
social space, and information space infiltrate each other. Educational environments
are transformed by the technology and interwoven with reality and seamless flow.
We are moving forward from the age of Internet and AI. Intelligence has become
an essential attribute of an educational environment. Intelligent educational envi-
ronments can automatically perceive the needs of teachers and students and provide
smart services.
Ubiquitous computing makes intelligent educational environments possible. The
idea of ubiquitous computing was initially elaborated by Mark Weiser. He advocated
that the use of computers is supposed to cater to people’s needs in daily life. He
believed that the development of ubiquitous computing will lead people into the era
of calm technology—technology is seamlessly integrated into our life, which would
free people from the sense of oppression, and makes it an unobstructed element [1].
Different from the previous computing-centered mode, pervasive computing pays
more attention to users and their satisfaction about self-experience. Ubiquitous
computing emphasizes the central role of people in the technical environment and
the effect of technology on people. Pervasive computing fully integrates physical
space and information space and mobilizes all available technologies to promote
human development. Mark Weiser describes the life of Sal in ubiquitous computing
as follows. When Sal gets up in the morning, the system has prepared coffee for
her. When Sal takes a look at the window, the system will tell her what is going on
around her. When Sal wants to know about current road conditions, the system will
send her real-time traffic information [2]. These scenarios demonstrate the features
of ubiquitous computing as universal, transparent, and adaptive.

30 3 Intelligent Learning Environments
As technology advances, the perception and computing power of the terminal

equipment is also getting stronger. It integrates more sensors, detectors, collector,
through which it can capture users, equipment, location, problems, and coping strate-
gies in the real world. It connects the virtual world with the real world and brings
more and more intelligent and personalized service so as to create an integrated
intelligent education environment.
When ubiquitous intelligent terminals and networks are integrated into educa-
tion, they will provide a brand-new ecological mechanism for education and teaching.
They will also enable a new human–machine relationship between learners and tech-
nology and fundamentally change the conception and approaches of learning. Ubiq-
uitous intelligent devices and computing power will establish a seamless connection
among people, objects, and machines—satisfying various possibilities of learning
at any time. Learning in an intelligent education environment will be ubiquitous,
namely U-learning.
U-learning emphasizes that anyone can obtain any learning resources to enjoy
ubiquitous learning at any time, in any place, with any computing devices. When we
encounter problems in daily life, we can rely on context-aware devices to perceive
requirements and send demand information to the cloud computing platform through
ubiquitous networks. The cloud computing platform will conduct seamless retrieval,
aggregation, calculation, and transformation to find the most suitable learning content
based on the needs of a user and personalized profile in the knowledge space. Learning
occurs everywhere. Learning, life, and work are integrated. And life itself is learning.
Ubiquitous learning in intelligent environments has the following features:
ubiquitous, continuous, social, situated, adaptive, and connected.
(1) Ubiquitous: Ubiquitous learning emphasizes the use of mobile and ubiquitous
technologies to meet learners’ demands for learning anytime and anywhere.
Learning services and resources are widely found in our living environment.
And learning opportunities are omnipresent and all-encompassing to enable
interactive learning anytime and anywhere.
(2) Continuous: Ubiquitous learning is a combination of formal learning and
informal learning, individual learning, community learning, and complement
classroom learning with online learning. Its continuity is manifested in two
aspects. On the one hand, it realizes the continuity of physical space. Learning is
integrated into work and daily life. Learners can seamlessly switch in multiple
contexts. On the other hand, the continuity of formal learning and informal
learning can support education on campus and continuing education at work.
Case 1 The MWOW Project

The Museum Without Walls (MWOW) Project [3], developed by the American Public
Art Association and Philadelphia Park and Entertainment, is supported by wire-
less handheld devices, embedded devices, and wireless networks. Learners enter a
location-based storytelling learning environment. Learners can choose where to go
or watch the art exhibits or school history of a museum of the art online. For example,
3 Intelligent Learning Environments 31
learners can choose MIT to browse the more than 100-year history of MIT through
devices such as mobile phones.
The MWOW Project has integrated technology with learners in an optimal so that
learners can focus on their learning. Learners can browse the Web to learn historical
figures or download an app that scans people or other works of art. For example, on
a particular tourist route, learners choose to browse the statue of Lincoln. The page
that opens is a text introduction to the creator of Lincoln as well as the audio and
video introduction.
(3) Social: Learning is the process of sharing and constructing individual cognitive
networks and social cognitive networks. Personal knowledge constitutes an
internal cognitive network. Situational learning resources in learning spaces
and other learners represent social cognitive networks. Learners refine their
cognitive networks in situated interaction. It also forms a part of the social
cognitive network at the same time to share and build the social cognitive
network.
(4) Situated: Situational awareness is the core element of ubiquitous learning.
Ubiquitous learning focuses on the interaction between physical and social
scenes and individuals. Different from constructivism, which plays an emphasis
on the construction of personal knowledge, ubiquitous learning turns the atten-
tion to the whole learning context. It believes that the situational network and
learning activities in which learners are located are the key to assist and support
learners to achieve learning goals.
(5) Adaptive: On-demand learning will become the main goal in ubiquitous
learning. Learners can obtain the content they need in the most appropriate way
of organization, expression, and service. Due to the different educational back-
ground, knowledge skills, and cognitive style of each learner, the intelligent
learning environment will provide a suitable individualized learning model
based on the learner’s historical learning data and push the corresponding
learning content to provide personalized learning service.
(6) Connectivity: Ubiquitous learning environments can develop learners’ ability
of independent learning in different situations and help learners to link and
integrate knowledge through understanding and processing. In the process of
interacting with the situation, learners complete the process of constructing
the individual cognitive network and the social cognitive network. The indi-
vidual cognitive network is composed of the learner’s personal knowledge and
the social cognitive network of different leaners’ situational problems in the
learning space. This process not only enables the construction of individual
cognitive network but also improves the social cognitive network and makes
the link from information to person possible.
Education is a systematic project. The educational environment is an organic
whole containing all the factors affecting the survival and development of the project.
The educational environment includes not only the classroom where learning takes
place but also the teaching management space. This chapter mainly introduces
the environments where education occurs, including smart classrooms, intelligent
library, intelligent campus security detection, and early warning. The application
scenarios of smart technology in constructing an intelligent education environment
are introduced in different cases in detail. From architectural sites to learning tools,
these cases show how intelligent technology transforms the way teachers and students
interact with learning resources to enable personalized learning. They also lay a
foundation for the construction of an intelligent educational environment in future
schools.
3.1 Smart Campus
The campus is the principal place where formal learning takes place. The construc-
tion of the campus directly affects the operating efficiency of the whole system.
Emerging technologies, like cloud computing, IoT, big data, mobile communica-
tions, AI, social networking, connected mobile campus, are widely used in higher
education. The relationship between information technology and education evolves
from combination and integration to fusion innovation. The smart campus is no
longer an idea in mind, slowly becoming a reality [4]. In general, a smart campus
is supported by an intelligent environment, resources, management, and decision-
making. On this basis, a variety of educational applications would be built so that
anyone can enjoy the learning services anytime and anywhere.
Smart campus is supposed to be personalized and service-oriented. An intelligent
campus can fully perceive physical environments and intelligently identify teacher
and students’ learning status and progress. It connects physical space and digital space
seamlessly to support the intelligent analysis of the learning process, assessment, and
decision-making [5].
The environment of the smart campus and all the functions are built based on
personalized service. The application of various vital technologies is aimed to effec-
tively meet teachers’ and students’ needs of life, learning, and work. Key technologies
supporting smart campus include learning scenario recognition and environmental
perception technology, connected mobile campus technology, and social network
analysis technology. Among them, context-awareness technology is the fundamental
technology of smart campus. Comprehensive perception in the smart campus consists
of two aspects. One is that the sensor can perceive, capture, and transmit information
about people, devices, and resources anytime anywhere. The other is to recognize,
capture, and delivery learners’ traits (learning preferences, cognitive characteristics,
attention status, learning styles, etc.) and learning scenarios (time, location, space,
learning partners, learning activities, etc.).
Environmental perception, for example, is enabled by installing sensors within the
campus. These sensors can sense people’s motion, light, sound, and temperature and
monitor the physical teaching environment (such as carbon dioxide levels in the air,
the temperature and humidity, the illumination, the air quality, and the noise level in
the campus.). With these sensors, the intelligent control systems could automatically
adjust the brightness of computer screens and the height of the curtain according to
3.1 Smart Campus 33
the real time environment in the classroom [6]. The comprehensive realization of
school teaching and intelligent management of a safe environment will create more
suitable conditions and a comfortable learning environment for students.
In the smart campus, valuable documents and items are affixed with QR codes
and GPS labels. With these codes and labels, the system can perceive and keep
track of their location. The function of positioning and monitoring vehicles entering
and exiting the campus can maintain the order of campus traffic. Installing intelligent
sensors in critical areas of the campus and necessary facilities is an important measure
to maintain campus security. These sensors can alert the management center and
analyze the cause of the incident if suspicious circumstances arise [7]. The integration
of RFID technology into teachers and students’ ID card, books, instruments, eleva-
tors, lamps, and other items enables personnel management of building access, class-
room and conference intelligent attendance, self-service borrowing and returning
booking, anti-theft and positioning of valuable equipment, laboratory control as well
as lighting, air conditioning and ventilation system control.
Case 1 Tencent Smart Campus

Tencent Smart Campus is the mobile ecosystem developed by Tencent for all levels of
schools and institutions. Tencent Smart Campus is integrated with Tencent’s WeChat,
News APP, Map, and other platforms, which supports student management, online
office, online training, online approval, and resource sharing in an intelligent terminal.
Tencent Smart Campus has five main functions [8]. An intelligent message can
convey the school’s notice teachers and students in real time. Students use WeChat
to share their ideas with the headmaster. Meanwhile, the intelligent message can
be integrated with the campus media to build a platform to share those messages
with the Tencent news client. Students can check the courses offered with one click.
They can also find the available room for learning by searching online. Moreover,
the initiation of campus activities is much simpler than that in traditional campus. It
also enables one-click payment to simplify the process of distributing subsidies and
scholarships. Parents can know the whereabouts of their kids and get their learning
status at every stage. It also provides a community for parents to communicate, which
makes school affairs communication more convenient.
Together with external high-quality educational resources, Tencent Smart Campus
provides schools, teachers, students, and parents a platform that integrates learning,
daily life, management, office business, and so on. With Tencent’s Smart Campus,
independent individuals are connected to an intelligent network.
Case 2 Smart Campus in Jiangnan University

Energy conservation and emission reduction is distinguished feature of the smart
campus of Jiangnan University in China. Since 2005, the new campus of Jiangnan
University has been put into use gradually. The number of teachers and students
studying and living in the new campus has increased significantly. Meanwhile,
the number of energy consumption equipment has increased substantially. To
address the issue of energy regulation, Jiangnan University aggregated experts from
different departments to design and develop an energy regulatory platform. This
platform provides better services to managers in their decision-making and effective

management [9].
As shown in Fig. 3.1, the platform adopts the model of centralized planning and
step-by-step implementation, using the “1 + 1 + N + M” architecture, which is
a platform, an energy service portal, a number of (N) business subsystems, and
mobile terminals. The module of regulation, control, and service is independent of
each other. The construction of relatively separate modules makes it convenient for
different groups of users. For example, the energy-saving authorities focus on regu-
lation. While other departments, like water and electricity maintenance and property
management, focus on “control” and service. With the help of nearly 20,000 types of
sensor monitoring points in the campus, real-time digital management of energy use,
water supply network, street lamps, security, and transportation becomes possible.
Over 90% of the whole campus is under monitoring [10]. The system provides
services like intelligent regulation, data statistics, and analysis of energy consump-
tion in all areas on campus. Managers can obtain comprehensive energy demand
statistics reports and analysis reports in the system. Besides, the platform can accu-
rately perceive and determine where energy consumption vulnerabilities exist to
better utilize resources and save costs to establish an orderly management model.
The power metering management platform regulates the energy usage of 1 million
square meters in the campus. The monitoring system of the water pipe network can
detect the leakage of underground pipes in time. The intelligent lighting control
system provides excellent management of all streetlights on campus [11]. Moreover,
the subsystem offers teachers and students their own energy consumption status.
Also, Jiangnan University has developed and released the mobile client to meet the
needs of users’ query of personal energy use as well as real-time monitoring.
Jiangnan University has produced significant economic benefits through the
construction of a digital energy regulatory platform. In 2017, the total amount of
equipment was 4.43 times that of 2006. The total number of air conditioners is
about 4.16 times the amount for 2006. However, the net hydropower expenditure is
merely 1.37 times the amount of 2006. And the performance of energy conservation
Fig. 3.1 Architecture of the digital energy regulatory platform

3.1 Smart Campus 35
and emission reduction is outstanding. Today, the annual expenditure of Jiangnan

University on water and electricity is controlled within 3% of the total daily expenses
(including personnel funds, public funds, scientific research funds, excluding infras-
tructure costs) [12]. The different levels of media have reported Jiangnan University’s
achievement of energy-saving and green campus building. It has provided advisory
services for more than 30 institutions [13].
The smart classroom is the second feature in Jiangnan University. The university
has made numerous attempts since 2012 to make the teacher-directed and student-
centered teaching mode possible. Specifically, Jiangnan University has established a
comprehensive renovation of the public areas of the teaching building and the interior
of the classroom. The school installs HD cameras in both the front and rear of all
classrooms. Besides, Jiangnan University has installed an access control system at all
classroom gates to assist intelligent management. All classrooms open automatically
at 6 AM and close at 10 PM. With the help of the classroom access system, the
regulation of students’ classroom attendance is further strengthened. Combined with
the digital information of learning, living, swiping cards, Internet access, and unified
identity authentication, the university can keep track of students’ learning trajectory
tracking system to enable excellent management of the whole process of education
[14].
3.2 Campus Safety Detection and Early Alarming
Campus safety is the basis of school education and the fundamental guarantee of
education reform and development. The Outline of China’s Long-Term Educational
Reform and Development Plan (2010–2020) states that it is necessary to maintain the
harmony and stability of the education system earnestly and deepen the development
of a safe campus. At the same time, the plan emphasizes that it is of great importance
to strengthening safety education and safety management [15]. At present, all sectors
of society are very concerned about campus safety. Campus accidents will directly
affect the healthy growth of young people and the harmony and stability of the whole
community. Different levels of educational systems should strive to provide a safe,
healthy, and stable learning environment for young people.
Campus safety management includes a safe internal and external environment,
implementing safety education activities, preventing the occurrence of criminal inci-
dents, perfecting the safety management system, and improving the accident disposal
procedures. In the past few years in China, campus accidents have significantly
increased, which have brought about painful lessons and irreparable losses. The
development of computer vision and robot technology makes it possible for AI to
maintain campus security.
This section introduces a variety of AI in the application of campus security detec-
tion and warning, including patrol robots, accident protection, and intelligent school
bus. For example, they can use face recognition through computer vision technology
to identify different visitors. Through image processing and other technologies, the
monitoring system automatically analyzes images to determine whether there are

potential risks to teachers and students. With the positioning technology and IoT
technology, we can monitor the status of the school bus to ensure its safety. These
cases all demonstrate that AI technologies can play an important role in campus
safety.
Case 1 Campus Patrol Security Robot

Campus patrol security robot is a multi-functional integrated system that inte-
grates environmental perception, route planning, dynamic decision-making, behavior
control, and alarm devices. On most campuses, the typical way is to recruit campus
security personnel and set up patrols, which takes a long time. As an essential member
of intelligent security, the security robot can replace more labor force and become a
necessary tool for industrial transformation [16].
Campus patrol robot can collect the facial information of people entering the
campus through the eye image acquisition equipment to identify his/her identity. If
the relevant profile is not detected, the system will inform the security personnel to
authenticate. Robots located at the entrance will also collect students “facial infor-
mation and compare them with the information in the repository”. The robots will
identify students “attendance and exit information to ensure the safety of students
while they are on campus” [17].
Security robots can operate throughout the campus. The patrol functions can
be divided into two modes, which are automatic patrol and remote-control patrol.
Figure 3.2 shows a security patrol robot on campus. Automatically, the robot is
capable of building maps and real-time positioning according to the environmental
condition/situation. The robot will perform patrol tasks based on a predetermined
route and scheduled time. When the robot encounters obstacles, the vision system
sends signals to the chassis control system, which makes a decision and corrects the
path for the robot to continue its work. If the robot runs out of power, it can find
a charging station on its own. In the remote-control mode, the image acquisition
and transmission system are used for real-time image acquisition and processing.
Security robots on the campus monitor the campus in real time with the four front
and rear mobile HD digital cameras. Images detected by security robots are sent
back to the control system to make a decision. If it identifies strangers entering the
campus, the system immediately sends information to the main control office and
runs the alarm system.
Case 2 Intelligent School Bus

In recent years, school bus accidents frequently happen in China. These accidents
arouse deep public concern about school bus safety. With the development of commu-
nication technology, especially the rapid growth of the Global Positioning System,
the monitoring system in school buses has been greatly improved recently. Based
on IoT, the monitoring system provides a secure guarantee for the school bus. Since
October 2016, the government of Xinyu, Jiangxi Province, has created a school
bus management platform. In the platform, videos are recorded inside the school
bus for teachers and parents’ monitoring. Currently, the management platform is
3.2 Campus Safety Detection and Early Alarming 37
Fig. 3.2 Example of a security robot on campus
installed in 225 buses from 170 schools in urban and rural areas throughout the city.
It is reported that with GPS technology in each school bus, all school buses can be
monitored through this unified platform [18].
IoT refers to the interconnection of all devices through RFID, GPS, and a variety
of sensors and modern communication technology. According to an agreed commu-
nication protocol, intelligent identification, positioning, tracking, monitoring, and
management of a network are enabled. IoT technology connects the school bus with
the background monitoring system in real time. Therefore, all information about the
vehicle and kids in them can be extracted in real time and made full utilization of.
With IoT technology, the monitoring system collects real-time information about
school buses, such as the location, running status, and route of the vehicle, for the
safety management system to refer to [19].
The system collects students’ behavior data and compares it with the model of
dangerous behavior in the database. When a risk is identified, the system will auto-
matically open the emergency alarm. The system will also generate the school bus
status report and create a real-time recording of the school bus. Besides, the system
will inform the administration department and parents about the location of students
through SMS alerts. Besides, the unique weather forecast broadcast equipped in the
vehicle can offer timely alarm when there are rainstorms, fog, debris flow, and other
natural disasters [20]. With advanced technologies, the current school bus provides a
better and safer service to students. The communication among school buses, oper-
ating schools, and regulatory departments is more reliable and stable, which effec-
tively prevents risks in the operation of school buses. Meanwhile, the supervision of
school buses could be strengthened.
Another random document with
no related content on Scribd:
The Project Gutenberg eBook of Tom Swift and
his great oil gusher
This ebook is for the use of anyone anywhere in the United States
and most other parts of the world at no cost and with almost no
restrictions whatsoever. You may copy it, give it away or re-use it
under the terms of the Project Gutenberg License included with this
ebook or online at www.gutenberg.org. If you are not located in the
United States, you will have to check the laws of the country where
you are located before using this eBook.
Title: Tom Swift and his great oil gusher

or, The treasure of Goby Farm
Author: Victor Appleton
Illustrator: Walter S. Rogers
Release date: January 13, 2024 [eBook #72705]
Language: English
Original publication: New York, NY: Grosset & Dunlap Publishers,

1924
Credits: Delphine Lettau, Cindy Beyer & the online Distributed

Proofreaders Canada team at http://www.pgdpcanada.net
*** START OF THE PROJECT GUTENBERG EBOOK TOM SWIFT

AND HIS GREAT OIL GUSHER ***
TOM SWIFT AND HIS
GREAT OIL GUSHER
OR
The Treasure of Goby Farm

By
VICTOR APPLETON
Author of “Tom Swift and His Motor Cycle,”
“Tom Swift Among the Diamond Makers,”
”Tom Swift and His Flying Boat,”
“The Moving Picture Boys Series,” etc.
ILLUSTRATED
NEW YORK
GROSSET & DUNLAP
PUBLISHERS
Made in the United States of America

THE OIL SHOT A HUNDRED FEET INTO THE AIR IN A HUGE
BLACK GEYSER.
CONTENTS
CHAPTER PAGE
I The...................................................................
Falling Airplane 1
II Almost
...................................................................
a Tragedy 11
III An...................................................................
Attractive Offer 21
IV Reckless
...................................................................
Driving 30
V A Close
...................................................................
Shave 39
VI Suspicious
...................................................................
Circumstances 49
VII Trouble
...................................................................
Brewing 60
VIII Through
...................................................................
the Window 71
IX The...................................................................
Midnight Prowler 80
X Caught
...................................................................
in the Act 88
XI A Rascal
...................................................................
Foiled 95
XII The...................................................................
New Drill 102
XIII A Night
...................................................................
of Triumph 108
XIV Kicked
...................................................................
for a Goal 114
XV Off...................................................................
for the Oil Fields 121
XVI On...................................................................
the Trail of Fortune 129
XVII Closing
...................................................................
the Deal 136
XVIII A Test
...................................................................
of Courage 146
XIX On...................................................................
the Verge 153
XX Capping
...................................................................
the Gusher 164
XXI The...................................................................
Inquisitive Stranger 171
XXII Rad...................................................................
Turns Up 177
XXIII A Dastardly
...................................................................
Plot 184
XXIV Fighting
...................................................................
the Fire 193
XXV Victory
................................................................... 201
TOM SWIFT AND HIS GREAT

OIL GUSHER
CHAPTER I
THE FALLING AIRPLANE
“A grand day for a spin in the air, Ned,” remarked Tom Swift, as
he stretched his arms and looked through the window of his office.
“What do you say? Come along and let the wind blow some of the
cobwebs out of your brain.”
“Get thee behind me, Satan,” replied Ned Newton, the young
financial manager of the Swift Construction Company. “I’ve got a
heap of work yet to do in checking up this last monthly statement.”
“That’ll keep,” said Tom. “You’ll find the figures waiting patiently
for you when you get back. I know you’re well ahead of your work,
anyway, and a whirl in the circumambient will do you good. You see,
I’m only thinking of you.”
“Yes, you are, you old hypocrite!” laughed Ned, who was about
the same age and on the warmest terms of friendship with his
talented young employer. “I’ve known for the last half hour that you
weren’t paying any attention to what I’ve been saying and that your
fingers were itching to get hold of the control lever of the
Hummingbird. Well, I suppose you’ve got to get it out of your system,
and it might as well be now as later. You go and tune up the old bus,
and I’ll be with you just as soon as I’ve given some directions to my
stenographer.”
Tom went off to the flying field, where several airplanes of his
own construction, from the smallest to the largest sizes, were kept in
readiness for use, and selected one which he had dubbed the
Hummingbird because of its diminutive size. Ned had suggested the
name, Aerial Mosquito, but Tom had rejected this as casting a stigma
on his pet which had never “stung” them yet.
It was an attractive little craft, and Tom patted it lovingly as he
went over every strut of the plane to see that it was in perfect
condition. Long experience in flying had taught him to take nothing
for granted. But there was nothing, even to his critical eye, that was
lacking, and the same was true of the engine, which purred smoothly
as it broke into its rhythmical song.
“Listen to that!” Tom cried enthusiastically to Ned, who by this
time had rejoined him. “Isn’t she humming sweetly? Never misses a
note. No grand opera prima donna has anything on her!”
“She sure sounds good,” agreed Ned, as he climbed into his seat
and adjusted his straps.
With one last look that took in everything, Tom followed him and
started the plane. The take-off was perfect. She ran along the
ground for a little distance and then soared into the air like a bird.
Tom let her climb, feeling out the air currents, until they were at a
height of about eight hundred feet. Then he put her on a level keel
and settled back in his seat to enjoy his ride to the full. At that
moment he would not have exchanged places with any one on earth.
It was one of the perfect days that come in the late spring. There
was scarcely a cloud in the sky. The sun was warm but not
oppressive, and was tempered by a light and refreshing breeze.
Below stretched a vast panorama of hill and valley and woodland,
the trees and meadows luxuriant with their new coat of green. At the
moment they were above Lake Carlopa, and they could see its
waters gleaming like a mirror of crystal beneath the rays of the sun.
“Some scenery!” exclaimed Tom jubilantly.
“You said it,” agreed Ned. “It reminds me of that iceberg where
we found the treasure——it’s so different.”
“Where we found the treasure but nearly lost our lives,” replied
Tom, his face growing a little sober with the recollection. “Phew! it
makes me shiver just to think of it. We’ve been in some pretty tight
places, Ned, but that was the closest squeak of all.”
“I wouldn’t care to repeat it, for a fact,” said Ned. “But I’ve no
doubt that you’ll be going into something just as dangerous before
long. You’re a glutton for danger. I’ll bet you’re just pining for some
other adventure.”
“Nothing in sight just now,” disclaimed Tom, with a laugh. “But I’m
not saying that if anything beckoned I’d give it the glassy stare.
There are lots of things we haven’t tried yet.”
“I’d like to know what they are,” returned Ned skeptically. “You’ve
been near death as many times as I have fingers and toes.”
“And yet you’ll notice that I’m far from being a dead one yet,” said
Tom lightly. “Or if I am dead, I’m a pretty lively corpse,” he added.
“Don’t worry about me, old man.”
“The pitcher that goes to the well too often, you know,” warned
Ned.
“Yes, I know that old wheeze,” laughed Tom. “But I’m not a
pitcher and——Great Scott! what’s that fellow up to?”
The exclamation was wrenched from him by the peculiar
gyrations of an airplane about two miles distant. They had noted its
presence some time previously, but as they were not far from an
aviation field and it was a common thing to see planes flying about,
they had not given it any special attention.
Now, however, they noted that the plane was behaving much as
a ship might which had lost its rudder. Its motions were confused and
erratic. It would plunge downward nose first and then right itself
abruptly and sail about in circles. There seemed to be no coherent
plan on the part of the aviator, and indeed it acted as it might if the
aviator were missing.
“Some fool aviator trying to do stunts,” was Ned’s comment, after
he had watched it for a moment. “Those fellows make me tired.
There are risks enough in flying without going out of the way to find
them.”
“No, I don’t think it’s that,” said Tom anxiously. “It looks to me as if
the pilot were in some sort of trouble. Lost control, sick, or
something. I’m going over to see.”
He turned the nose of his craft directly toward the strange plane
and made for it at full speed.
The queer contortions of the plane still persisted, and Ned was by
this time convinced that Tom was right. A thrill of horror went through
them as the conviction grew upon them that they might be about to
witness a tragedy. If the plane should dash to the ground from such
a height, there was not one chance in a thousand that the luckless
pilot would survive.
They could see by this time that there was only one man in the
plane, which was a much larger craft than their own. They could see
him working desperately to get his craft under control. Where the
trouble lay, whether with the plane or the engine, they had no means
of telling. But that the man was in deadly danger and knew it, there
was no longer any room for doubt.
They had come within half a mile of the strange airship when the
disaster they had been dreading happened. The plane suddenly
turned its nose to the earth until it pointed directly downward. Then it
fell like a plummet, down, down, into the midst of a forest, where it
landed with a terrific crash and was lost to sight.
A cry of horror burst from Tom and Ned.
“Quick, Tom!” cried the latter. “Can’t you make a landing
somewhere?”
“That’s what I’m looking for,” replied Tom, as his eyes swept the
surrounding country in search of some open spot.
But he looked in vain. The woods extended for several miles in
every direction. Here and there were small open places, but as Tom
circled about, flying as low as he could without striking the tops of
the trees, he saw that none of them was large enough to permit the
landing of the plane.
“There’s but one thing to do,” he said in a tense tone, when this
conviction had become certainty. “We’ve got to mark the location of
this place so that we can find it when we come to it again. I’ll leave
you to fix that in your mind. Now we’ll make straight for home, get
out my roadster and try to get to the place as soon as possible. The
chances are that the poor fellow is dead, but we’ll do all we can. You
get busy on the radio, so that the car will be all ready for us to jump
into the minute we get back to the works.”
Ned turned his attention to the radio set with which the plane was
equipped and called up Garret Jackson, the shop manager for the
Swift Construction Company. He told of the accident and directed
Jackson to have Tom’s roadster ready and the engine going.
Tom in the meantime had been making the Hummingbird travel at
her fastest speed, and only a few minutes elapsed before they
reached the yards. He made a perfect landing, and before the plane
had fairly stopped he and Ned had leaped from it and were running
toward the smart roadster, alongside of which Jackson was standing.
“I’ve put the first-aid kit under the seat in case you should need
it,” said Jackson.
“Good work, Jackson,” commented the young inventor, as he
jumped in and took the wheel. “Perhaps you’d better come along
with us. There may be work for all of us. All set? Let’s go.”
He threw in the clutch and the speedy machine started and was
soon racing along at a record-breaking speed in the direction of the
woods. Although they were fairly flying, it seemed to Tom and Ned in
their anxiety that they were crawling.
“Give her the gas, Tom,” urged Ned.
“She’s making sixty now,” replied Tom, with a glance at the
speedometer. “Got the exact location, Ned?”
“Pretty nearly,” replied Ned. “Turn to the right at the first road after
we pass the church at the forks. That’ll bring us to the Thaxton
woods in which the plane fell. There’s an old woods road that runs
through the center of the forest, and if we keep our eyes open on
either side we’ll probably find it.”
Soon they had passed the church and turned into the road that
Ned had indicated. At that point the going was good, but as they
advanced it gradually grew more rutted and rocky. Tom kept up a
high speed, however, though they had to hold on and the car
bounded up and down in a manner that threatened wreck if the pace
were maintained.
But a human life was in the balance, and if that could be saved
nothing else mattered.
At last, deep in the woods, Tom slackened speed, for the road
had “petered out.” But he could see that where it ceased to be a road
it had become a trail that wound its uneven way among the trees. It
was barely wide enough for the car to pass along, and here and
there in the road were fragments of stumps that made traveling
precarious.
“Gee, Tom, you’ll have to go slow here,” declared Ned.
“We don’t want to break our necks,” added Jackson.
“We’ll do the best we can,” answered the young inventor. “But if
either of you wants to get out and walk——”
“Nothing doing, Tom!” cried his financial manager quickly. “If you
stick, so do I.”
“Same here,” added the shop foreman.
On and on they rushed. Once they hit a big rock and for the
moment it looked as if the machine would surely go over. Ned and
Jackson held their breath. Maybe one or the other thought the next
minute would be his last.
But Tom was a skilled driver, and with almost miraculous deftness
he piloted the car at as fast a rate as he dared. He could not take his
eyes from the road, but Ned scanned the woods on one side and
Jackson on the other for any sign of the hapless airman.
They had gone perhaps half a mile when a shout came from the
lips of Jackson.
“The woods are afire!” he cried, as he pointed ahead of them.
They looked in the direction indicated and saw a column of
flames mounting above the tree tops.
CHAPTER II
ALMOST A TRAGEDY
Tom Swift and the two with him stared for a moment in
consternation.
“It isn’t the woods on fire,” cried Tom. “They’re too green for that;
and besides we had a drenching storm two days ago. It’s the plane
that’s blazing.”
“And that poor fellow perhaps is being burned to death!” cried
Ned. “Hurry, Tom; hurry!”
There was no need for urging. Tom threw caution to the winds
and tore along at a reckless pace in the direction of the flames.
As he turned a bend in the trail, an exclamation escaped him.
Right in the road was a small tree which had been blown down. It lay
directly in front of him, completely blocking the road.
He had scarcely time to stop before colliding with it. Then with
one accord all leaped from the car and tugged desperately at the
tree to try to move it to one side far enough to permit the car to pass.
Although they put forth almost superhuman efforts, they soon
found that their strength was unequal to the task. Tom was the first to
realize it.
“No use, fellows,” he panted. “We’ll have to leave the car here
and make the rest of the way on foot.”
They clambered over the tree and started running with all their
might in the direction of the blazing plane.
And while, with their hearts full of apprehension and their lungs
strained almost to bursting, they are hastening to the rescue, it may
be well for the benefit of those who have not read the preceding
volumes of this series to tell who Tom Swift was and what had been
his adventures up to the time this story opens.
Tom, now about twenty-one years old, lived in the town of
Shopton, a small inland city on the shores of Lake Carlopa. His
father was Barton Swift, an inventor of note. Tom had grown up
among wholesome surroundings and developed into a clear-cut,
athletic young fellow, a general favorite among his associates. He
was frank, courageous and good-natured, never looking for trouble
but never stepping aside when it turned up.
Tom may have inherited the inventive genius of his father, or it
may have been the force of example, but from his earliest years he
displayed a marvelous aptitude in all things scientific and
mechanical. A mere glance at any mechanical contrivance, from a
watch to a steam engine, generally sufficed to give him a pretty clear
idea of its working. But this was not enough. He saw not only its
good qualities, but also its defects, and his active mind busied itself
at once in trying to devise improvements. His success in this had
been phenomenal. From his first venture into the realm of invention
described in the opening volume of the series, entitled “Tom, Swift
and His Motor Cycle,” his progress had been rapid. His ambition
grew from what it fed upon, and the improvements that he added to
the motor cycle were duplicated by those described in later volumes
bearing on airships, submarines, rifles, cameras, searchlights,
cannon and a host of other things that became famous. His friends
described him as a “wizard.” But Tom, who was as modest as he
was ingenious, never thought of himself in that light. He was
unspoiled by praise and kept on devoting himself to his chosen
vocation. Like Alexander, he was always looking for new worlds to
conquer.
The work that he had done naturally brought with it large financial
rewards. His activities resulted in the formation of the Swift
Construction Company, which, starting on a small scale, had
attained great proportions and formed the leading industry of the
town of Shopton. In the development of the business Tom had been
greatly aided by the financial ability of Ned Newton, his closest
friend. Ned had been a clerk in one of the Shopton banks, but had
left this position to become treasurer of the Swift concern. He was a
genius in business management, and Tom was able with an easy
mind to leave all financial details to Ned while he devoted himself to
his inventions. It was an ideal combination. They were fortunate, too,
in having at the head of the mechanical department Garret Jackson,
a thorough mechanic himself and skilled in handling the large body
of workmen on the Swift payroll.
Tom did not care overmuch for money, but of course he was glad
that the business was prospering, especially as at the present time
he was deeply interested in a charming young lady, Miss Mary
Nestor. The attachment between them had grown rapidly of late and
had been increased on Mary’s part by a great service that Tom had
recently done in rescuing her father from an iceberg, on which he
had been marooned while in a search for health. What peril was
involved in that rescue and how Tom’s courage and ingenuity
prevailed, are thrillingly told in the preceding volume of this series
entitled: “Tom Swift and His Flying Boat; or, The Castaways of the
Giant Iceberg.”
Now to return to the three who were hastening in the direction of
the burning plane.
It must be confessed that they had little hope of being of use. The
chances were great that the aviator had not survived the fall. Still,
there was a chance, and that thought added wings to their feet as
they tore through the forest.
Tom was first on the scene, with Ned close at his heels and
Jackson bringing up the rear.
Resting on two trees was most of what remained of the doomed
plane. It was still blazing so fiercely that they could not approach it
closely. The motor had fallen through the trees to the ground,
together with part of the wreckage of the plane.
There was nothing to be seen of the aviator, and they scanned
the ground fearfully. Their first thought was that he must be lying
crushed under the heavy motor. That fate, tragic as it was, would be
better than being burned to death.
Suddenly, Tom, who had been examining all the trees that stood
outlined in that great zone of light, gave a startled exclamation.
“There he is!” he cried, as he pointed to the branches of a tree
fully fifty feet distant. “There, in the crotch of those two big
branches.”
The others looked and saw the body of a man at the place
indicated, about twenty feet from the ground. His torn clothing had
caught in the branches, and big boughs supported the greater part of
his weight. There was no sign of life as they gazed at the crumpled
mass.
“He’s been thrown or jumped from the plane!” exclaimed Tom, as
all three rushed to the tree. “Here, Ned, give me your back and we’ll
get him down.”
Ned bent his back, and Tom, with Jackson’s assistance, got up
on it. Even then, with his six feet of height and his long reach, he
found that he was several feet short of the lowest branch.
“Hold steady, Ned, I’m going to jump,” he warned.
Ned braced himself, and Tom gave a mighty leap, just clutching
the bough with his out-stretched hands. It was enough, however, and
he swung his legs up, and in a moment was in a sitting position.
“Now, Ned, you do the same thing,” he directed. “I won’t be able
to bring him down alone. You stand by, Jackson, when we lower him.
In the meantime run back to the car and get the coil of rope that’s
under the seat. And bring the first-aid kit along with you at the same
time.”
Jackson waited only long enough to enable Ned to get up in the
same way Tom had done, and then set off on a run to the car.
Tom and Ned made their way up through the branches until they
were close to the unfortunate victim of the disaster. They had to
proceed with caution for fear the swaying of the branches might
release the body and let it fall to the ground. But at last they were in
a position where they could see the man’s face.
It was a youthful face, that of a young man apparently no older
than themselves. It was badly scratched and bruised, and blood was
flowing from several wounds.
Tom reached under the torn coat and placed his hand on the
man’s heart.
“Still beating!” he exclaimed joyfully. “Not very strongly, but he’s
alive and he’s got a fighting chance. Now, Ned, lend a hand and we’ll
get him down from here in a jiffy.”
That, however, was more easily said than done. It was a difficult
thing to release the body from the twigs and branches that held the
clothing. Then, too, each of them could use only one hand to support
the weight of the body, while with the other they must cling for
support to the branches. It was a task that taxed their strength to the
utmost.
Luckily they were both stalwart and vigorous and in superb
physical condition, and the fearful extremity in which their helpless
burden lay called into play the reserve strength that can usually be
depended on to meet a great emergency. Slowly and painfully they
made their tortuous way down among the branches, until they rested
on the lowest bough, one that seemed strong enough to support a
regiment.
They paused there a moment to take breath. Jackson, in the
meantime, had returned with a stout coil of rope. He tossed one end
of it up to Tom. The latter caught it and made a loop which he
fastened securely under the arms of the unconscious youth. Then,
with Ned also holding on, he slowly paid it out until the aviator was
within Jackson’s reach. The moment this was done, Tom and Ned
dropped to the ground, and together they unfastened the rope and
laid the airman gently on the grass.
As they did so, the youth gave a groan and opened his eyes. He
looked vacantly at the faces above him and then drifted again into
unconsciousness.
There was a brook at a little distance, and they carried him to the
bank. Then, while Ned and Jackson bathed his face and wrists with
the cool water, Tom ran his hands over the body. He had had a good
deal of experience with accidents, and this now stood him in stead.
His face was grave when he had finished.
“His right leg is broken, and a couple of ribs as well,” he
announced. “He may have internal injuries also, but it will take a
doctor to tell that. Hand me over that first-aid kit and I’ll make a splint
for the leg. Then we’ll get him to the hospital as soon as possible.”
Tom worked skillfully and rapidly and soon had the temporary
splint in place. The pain involved in setting the broken bone roused
the wounded aviator from his oblivion. Again he opened his eyes, but
this time there was less of vacancy in them. He tried to speak, but
though his lips moved, no sound issued from them.
Tom bent over him.
“You’ve been hurt,” he said gently. “But you’re with friends. If you
tell us what your name is and where you live we’ll try to get in touch
with your people.”
The young man muttered something that was unintelligible.

Ebook An Introduction To Artificial Intelligence in Education Yu Online PDF All Chapter

Uploaded by

Copyright:

Available Formats

Ebook An Introduction To Artificial Intelligence in Education Yu Online PDF All Chapter

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Ebook An Introduction To Artificial Intelligence in Education Yu Online PDF All Chapter

Uploaded by

Copyright:

Available Formats

An Introduction to Artificial Intelligence

Artificial Intelligence With an Introduction to Machine

Introduction to Algorithmic Marketing Artificial

Artificial Intelligence in STEM Education : The

A Machine Learning Artificial Intelligence Approach to

An Introduction to Education 1st Edition Hasan Arslan

Neural Assemblies An Alternative Approach to Classical

Cognitive Electronic Warfare: An Artificial

Swarm Intelligence An Approach from Natural to

More information about this series at https://link.springer.com/bookseries/16427

ISSN 2662-5342 ISSN 2662-5350 (electronic)

© Springer Nature Singapore Pte Ltd. 2021

assessment, intelligent support for teachers, and AI-supported education manage-

Beijing, China Shengquan Yu

5 Intelligent Teacher Assistant . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101

© Springer Nature Singapore Pte Ltd. 2021 1

1.1 Cognition and Learning in the Age of AI

Intelligence is the intellectual capacity of humans, including perception, memory,

1.2 A Brief History of AI in Education

AI technologies have a long history of supporting effective learning. The Intelli-

1.3 Typical Areas of AI in Education

AI in education is a basic approach to realize future education and a fundamental

Fig. 1.1 Key aspects of AI in education

capacity structures, intelligent diagnosis and recommendation, cognitive load moni-

1. China Institute of Electronic Technology Standardization (2018) Artificial intelligence stan-

2.1 Basic Approaches in AI

© Springer Nature Singapore Pte Ltd. 2021 9

two-dimensional image recognition with distorted invariances, such as image recog-

2.2 Major AI Technologies

computer vision, human–computer interaction, virtual reality and augmented reality,

A General Framework of AI in Education

Fig. 2.1 General technology framework of AI in education

educational materials, such as textbook learning, tutoring exercises, and instruction

The ultimate goal of applying AI in education is to improve learning efficiency and

Cloud Computing and Big Data

Fig. 2.2 5 V features of big

Fig. 2.3 Technology framework of machine learning

Supervised Learning: Supervised learning refers to the use of a given labeled

Natural Language Processing

Fig. 2.4 Technology framework of NLP

Robot and Intelligent Control

different fields. In the field of education, there is an educational robot developed by

Virtual Reality and Augmented Reality

2.3 Typical Open-Source AI Projects and Cloud Services

In addition to abundant algorithm services, PaddlePaddle is featured by its high

Tencent AI Open Platform

iFLYTEK Open Platform

Amazon Web Services

Currently, intelligent terminals are gradually becoming popular. Smart terminals,

© Springer Nature Singapore Pte Ltd. 2021 29

As technology advances, the perception and computing power of the terminal

Case 1 The MWOW Project

3.1 Smart Campus

Case 1 Tencent Smart Campus

Case 2 Smart Campus in Jiangnan University

platform provides better services to managers in their decision-making and effective

Fig. 3.1 Architecture of the digital energy regulatory platform

and emission reduction is outstanding. Today, the annual expenditure of Jiangnan

3.2 Campus Safety Detection and Early Alarming