Case Study

Effective Risk Positioning through Automated

Identification of Missing Contract Conditions from the
Contractor’s Perspective Based on FIDIC Contract Cases
JeeHee Lee, Ph.D. 1; Youngjib Ham, Ph.D., A.M.ASCE 2; June-Seong Yi, Ph.D. 3; and JeongWook Son, Ph.D. 4
Downloaded from ascelibrary.org by University College London on 03/01/20. Copyright ASCE. For personal use only; all rights reserved.

Abstract: Defining, measuring, and dealing with contractual risks are crucial for successful construction projects because the contractual
risks can lead to serious claims and disputes. In general, construction participants make a stipulation regarding their roles and responsibilities
by contracting in order to prevent such claims and disputes. A common practice for preparing construction contracts is to modify the standard
contract forms to reflect the interests of the given project from the owner’s perspective. In this process, however, favorable clauses that may be
beneficial to the contractor are often modified or even removed, causing significant potential risks to the contractor. Therefore, an in-depth
review of contract terms and conditions is required to avoid future risks. This study presents a new proactive risk assessment model to identify
missing contractor-friendly clauses in the owner’s modified contract conditions from the contractor’s point of view. A case study is used to
demonstrate the proposed framework, and real-world project cases were analyzed to understand what type of contractor-friendly clauses
would likely be omitted in the owner’s modified contract. In this study, the developed model builds on rule-based natural-language processing
(NLP) to analyze unstructured text data through preprocessing, syntactic analysis, and semantic analysis. The proposed data-driven risk
assessment model is expected to reduce the extent of human errors by (1) identifying potential contractual risks that could arise disputes;
and (2) supporting to develop an appropriate response strategy for the given risks. DOI: 10.1061/(ASCE)ME.1943-5479.0000757. © 2020
American Society of Civil Engineers.
Author keywords: Contract conditions; Contractor-friendly clauses; Contract risk; Unstructured text data; Rule-based natural-language
processing (NLP).

Introduction all participants in construction projects typically make a stipulation

regarding their roles and responsibilities by contracting.
Given the nature of construction projects, claims and disputes A construction contract is a mutual written agreement signed by
among various stakeholders are inevitable, which often result in contracting parties (Gao et al. 2018; Zaghloul and Hartman 2003).
cost overruns and schedule delays (Awwad et al. 2016; Cakmak The contract is a dominant formal governance mechanism for
and Cakmak 2013; Lee et al. 2019; Li and On Cheung 2018; resolving the disputes across stakeholders, and the contractual facts
Youssef et al. 2018). It has been reported that claims and disputes set forth in the contract serve as the basis for important judgments
not only hamper the execution of successful projects, but they can when disputes arise (Gao et al. 2018; Lee et al. 2019; On Cheung
also lead to massive cost losses, which can reduce the profitability et al. 2018). In particular, terms and conditions of contracts address
of the construction entity (Cheung 1999; Niu et al. 2015; Semple large portions of risk factors, and the contract terms determine
et al. 1994; Sertyesilisik 2010). For example, according to a recent whether the contracts or its clauses are risky or not to either party
report by Arcadis (2018), it was observed that the average cost of of the contract (Youssef et al. 2018). Therefore, the obligations and
dispute settlement in the global construction market in 2017 was rights of each entity should be clearly stated in the contract, and the
about $43 million, and the average time to resolve a dispute terms and expressions of the contract should be carefully examined.
was 14.8 months. To prevent such construction claims and disputes, In the construction industry, standard forms of contract condi-
tions [e.g., the International Federation of Consulting Engineers
1 (FIDIC), Joint Contracts Tribunal (JCT), and the American Institute
Postdoctoral Research Fellow, Dept. of Construction Science, Texas
A&M Univ., 3137 TAMU, College Station, TX 77843. Email: jhlee04@
of Architects (AIA)] have been widely used between owners and
tamu.edu contractors because these can reduce the possibility of misunder-
2
Assistant Professor, Dept. of Construction Science, Texas A&M Univ., standings, the likelihood of change orders, and the occurrence of
3137 TAMU, College Station, TX 77843. ORCID: https://orcid.org/0000 disputes (Bubshait and Almohawis 1994). In particular, the FIDIC
-0001-7157-4878. Email: [email protected] standard contract conditions are among the most globally recog-
3
Professor, Dept. of Architectural and Urban Systems Engineering, nized and used in international construction projects including
Ewha Womans Univ., 52 Ewhayeodae-gil, Seodaemun-gu, Seoul 03760, those Europe, South America, Asia, Africa, and the Middle East,
Korea (corresponding author). Email: [email protected] with a balanced approach to the roles and responsibilities of the
4
Associate Professor, Dept. of Architectural and Urban Systems Engi- contracting parties (Chen et al. 2013).
neering, Ewha Womans Univ., 52 Ewhayeodae-gil, Seodaemun-gu, Seoul
However, project owners often customize the standard contract
03760, Korea. Email: [email protected]
Note. This manuscript was submitted on May 12, 2019; approved on conditions by adding particular conditions from their perspectives,
October 8, 2019; published online on February 29, 2020. Discussion period which causes potential issues (Hamie and Abdul-Malak 2018b;
open until July 29, 2020; separate discussions must be submitted for indi- Rameezdeen and Rodrigo 2014). In the process of modifying
vidual papers. This paper is part of the Journal of Management in Engi- standard contract conditions, favorable clauses that are likely
neering, © ASCE, ISSN 0742-597X. beneficial to contractors are often modified or even removed.

© ASCE 05020003-1 J. Manage. Eng.

J. Manage. Eng., 2020, 36(3): 05020003

 Youssef et al. (2018) pointed out that owners tend to modify the likely be omitted in the owner’s modified contract. In particular,
general conditions in order to place more risks on contractors. Recent because the FIDIC contract conditions have become widely ac-
studies (Chen et al. 2013; Lee et al. 2019; Rameezdeen and Rodrigo cepted by employers and contractors as forming a reasonable and
2014; Youssef et al. 2018) have demonstrated problems that arise fair starting point in allocating risk between the parties (Butcher
from owner’s modified conditions of contracts (e.g., adding poison- et al. 2018), omission of favorable contract clauses away from
ous clauses and/or using ambiguous terms), which would most likely the FIDIC conditions likely to be potential issues. Therefore, the
cause contractor risks. In the practice of construction law, issues over missing clauses in this study can be defined as contractor-friendly
the interpretation of contract terms or clauses represent the main clauses omitted from the FIDIC standard conditions.
cause of litigation and arbitration, often because the parties have The proposed research is expected to contribute to the manage-
not appropriately drafted such contract conditions (Milner 2011). ment in engineering body of knowledge by enhancing domain-
In particular, disappearing important contract terms or clauses specific contract management skills to identify potential contractual
during the owner’s contract modification can cause significant risks that could arise disputes. Furthermore, this study ultimately
conflicts or disputes (Eggleston 2009; Sertyesilisik 2010). If the supports proactive risk management in the bidding and contracting
Downloaded from ascelibrary.org by University College London on 03/01/20. Copyright ASCE. For personal use only; all rights reserved.

disappeared contract terms are associated with guaranteeing any phase of construction projects by developing a quick and robust
party’s contractual rights, the level of risks can be particularly decision support tool for contractors; thus, the advanced under-
high. For example, the FIDIC conditions of contract specifies standing of management in engineering knowledge will help to
the limitations of liability to prevent malicious repetitive rework avoid uncertainties when drafting contracts.
or unlimited liability to contractors. However, if this provision is
omitted in the owner’s modified contract, it might cause potential
disputes that require the contractor to pay a huge amount of com- Related Works
pensation (sometimes more than the total contract amount) (ASCE
Insurance in Contracts 2015), which is a significant risk for the
Contract-Related Risks and Modified Contract
contractor. Because the limitation of liability clause is a contractual
Conditions from Owner’s Perspectives
ceiling on the amount of damages to be awarded if a plaintiff pre-
vails in later litigation between the contracting parties, a limitation Previous studies have demonstrated that the analysis and interpre-
of liability clause is a necessary precautionary piece of armor in tation of construction contracts is essential for preventing possible
contracts (ASCE Insurance in Contracts 2015). In this sense, con- future conflicts and disputes. Hamie and Abdul-Malak (2018a)
tractors should take a careful look at whether or not the limitation of highlighted the types of defects likely to be encountered in the con-
liability clause has been omitted during a contract drafting in order tract documents and showed that some of standard contract con-
to limit their liability in the event of a dispute. ditions were found to be tackling interpretation-related issues
Unfortunately, several studies (Lee 2018; Walsh 2017) have (e.g., terminology ambiguity, errors, and inconsistency). The Middle
shown that even experienced contract experts often overlook poten- East (ME), which is one of the growing construction markets in the
tial contract risks associated with missing terms and clauses. Ac- world, has witnessed many construction disputes because of amend-
cording to the results of a previous study (Lee 2018), it was ing standard contract conditions to shift most of the risks to the con-
observed that although poisonous clauses in the modified contracts tractor (Awwad et al. 2016).
were frequently found, the missing contractor-friendly clauses were Abotaleb et al. (2019) provided contract administration guide-
infrequently found in the contract review process. Lee (2018)’s re- lines for enhancing understanding of the contract parties to a proj-
search showed that only 54% of missing clauses were identified by ect’s existing contractual clauses related to the owner’s obligations.
professionals, whereas 82% of poisonous clauses were identified. They analyzed the provisions associated with the owners’ nonpay-
This indicates that the contractors likely make more mistakes in ment obligations under the most widely used international design-
finding missing clauses than finding problematic clauses in the con- build (DB) contracts (e.g., FIDIC and AIA) and highlighted the
tracts. Therefore, minimizing any mistakes that may occur in the differences and commonalities to provide an extensive checklist
contract review process should be considered as a top priority be- as a guideline for drafting contractual clauses. In order to inves-
cause signing the contract without being aware of risk factors may tigate the role of contract in coping with complexity-related risk,
cause a significant problem in the future. For that purpose, there is Gao et al. (2018) analyzed the efficacy of contractual functions
a critical need to minimize human errors and to avoid contractually under a diverse type of project complexity. The empirical results
unfavorable situations of contractors through a systematic approach showed that the contractual coordination can deal with risks in-
to identify risks associated with missing contractor-friendly clauses. duced by technical, organizational, and environmental complex-
The gap in satisfying such a need allows contractors to make stra- ity, whereas contractual control, such as control clauses involving
tegic decision making while reducing mistakes, especially for those guarantees, control rights, early termination, and sanction rights,
who need to submit a bid within tight time frames. is ineffective when either technical or environmental complexity
The objective of the study is to support identifying missing is high.
clauses in the owner’s modified contract conditions through a sys- In particular, several studies (Laryea 2011; Lu et al. 2016;
tematic approach in an automated manner. The challenge is to pro- Maemura et al. 2018) have emphasized the importance of the
vide a systematic support model that identifies the missing clauses clarity in contract terms because the ambiguity of contract terms
by better enabling computers to perform such knowledge-inspired may cause misunderstandings and conflicts among contract parties.
risk identification within a short period of bidding preparation time. Lu et al. (2016) presented that contract completeness, such as term
As a proof of concept, this study uses case studies based on 17 real- specificity and contractual obligatoriness, has negative effects on
world project’s contract documents modified away from the FIDIC contractors’ opportunistic behavior. Chong and Zin (2010) exam-
standard conditions. For each of the 17 case projects, the authors ined a standard contract form in Malaysia in terms of contractual
performed interviews with contract management experts and col- clarity. They identified 11 clarity problems and 4 legalese problems
lected project documentation for reviewing contract conditions, of the standard form of contract by questionnaire survey. If alloca-
and these data were then compared with FIIDIC standard condi- tion of risks between the parties is not sufficiently fair and balanced
tions to explore what type of contractor-friendly clauses would during the modification of standard contract conditions, it is likely

© ASCE 05020003-2 J. Manage. Eng.

J. Manage. Eng., 2020, 36(3): 05020003

 to lead to arising contractual issues. Hughes et al. (2015) pointed whether contractor-friendly clauses are missing or not in the con-
out that as contractors’ contract-management tasks become increas- tracts. Thus, NLP, involving lexical analysis, syntactic analysis, and
ingly common and specialized, owners are prone to maliciously semantic analysis, was employed for this research. Given the chal-
modify the conditions of contract, thereby shifting the risks to lenges that every contract has different structures and various
contractors. vocabularies (Nguyen et al. 2018), a robust content analysis ap-
According to Lee (2018), the main risks caused by owner- proach is needed, and thus sentence-level content analysis rather
modified contract conditions are whether (1) any hazardous clauses than word-level analysis alone was performed to conduct in-depth
are included in the contract, and (2) any required clauses are miss- semantic analysis. The analysis process of this study includes iden-
ing from the contract. Several studies have given useful insights tifying contractor-friendly clauses, determining whether the favor-
into the problems of owner-modified contract documents. For ex- able clauses are missing, and recognizing the type of risk in a clause
ample, Rameezdeen and Rodrigo (2014) highlighted the readability that is missing from the contract.
problems due to ill-modified contract conditions. Youssef et al. The purpose of using NLP in this study is to judge whether
(2018) focused on semantic issues associated with the allocation of the contractor-friendly clauses are omitted or not by carrying
Downloaded from ascelibrary.org by University College London on 03/01/20. Copyright ASCE. For personal use only; all rights reserved.

obligations and risks to a contract party. In short, previous research out syntactic analysis and semantic analysis on the texts of con-
presented that owner’s modified contract conditions transferring struction contracts. Natural language is composed of a collection
potential risks to the contractor should be managed in the contract of words called a lexicon and sentence structure rules called gram-
management aspect; however, systematic approaches for managing mar, and these are very complicated because they can produce vari-
contractual risks have not been applied to date. ous meanings according to the different combinations of words
used (Manning et al. 1999). Given the relatively limited vocabulary
Systematic Approach to Address Contract Risks used in a construction contract domain, the research challenge is to
maximize the effect of NLP based on domain-specific lexicon.
Contractors can ensure contract fairness by paying close attention This case study builds on international construction project
to what they are taking responsibility for and by including specific cases of Asia and Middle East in the process of model implemen-
remedies in the contract language for things beyond their control tation. The case projects are building and infrastructure construc-
(Craig 2016); however, in-depth analysis on large volume of con- tion projects, and the contract conditions are all based on the FIDIC
tract conditions is usually not easy within a short-term contract standard form. The contract conditions of the case projects were
preparation period. In order to address the challenges in reviewing modified away from the FIDIC standard form in terms of the own-
contract clauses, there have been few research efforts. Given the er’s perspective with a lack of the contractor-friendly clauses. Com-
fact that contract documents are expressed in natural language, paring the case projects with the FIDIC conditions allows for
natural-language processing (NLP) to extract a meaningful knowl- determination of which types of contractor-friendly clauses are fre-
edge from free text has been applied for analyzing and monitoring quently missing in the owner’s modified contracts. This study is
problematic clauses in the contract document. NLP involves lexi- conducted in the following four steps (Fig. 1), where in the figure,
cal, syntactic, and semantic analysis depending on the depth and
Step 1 shows a preliminary study collecting and categorizing case
level of analysis. Lexical analysis refers to word-level analysis,
data, and the case data collected are applied for natural language
and syntactic analysis refers to the analysis of sequences of words,
processing throughout Steps 2–4:
mainly sentence-level analysis (Lee et al. 2019).
• Step 1. Identifying contractor-friendly clauses: contractor-
For example, Al Qady and Kandil (2010) have developed a sys-
friendly clauses, which frequently disappear from the owner’s
tem that can extract semantic knowledge (i.e., concept relation of
modified construction contract, are defined based on the 17
contract conditions) from construction contract documents using
real-world projects’ contract documents and previous dispute
NLP. They identified three components of a concept set, automati-
cases (i.e., legal cases). These clauses are categorized as types
cally extracted the concept sets, and then evaluated the extraction
of contract risks (e.g., assignment, variation, and payment,
results by comparison with human evaluators. Lee et al. (2019) pro-
among others).
posed an automatic contract risk extraction model based on NLP.
• Step 2. Text preprocessing using simplification rules: simplifica-
The model detects poisonous clauses that are unfavorably modified
tion rules are developed in order to make compound and complex
from the FIDIC, the standard forms of contract, and gives warning
structures of sentences (e.g., parallel structures with conjunctions
to the contractors. Marzouk and Enaba (2019) applied text mining
and structures including several subordinate clauses) in the con-
based on NLP to identify project parties’ contractual obligations by
tract document concise to simple sentences with an independent
reducing time and effort needed for contract analysis.
clause.
Although considerable studies have been undertaken to reduce
• Step 3. Subject, verb, and object (SVO) tuple extraction: risk
risks underlying contract conditions in the construction sector,
identification elements (i.e., SVO tuple) are extracted in the pro-
there are still research gaps to supporting identification of missing
cess of syntactic analysis, and additional syntactic rules are de-
clauses in the owner’s customized contract conditions. Contract
veloped for extracting the SVO tuple in an exceptional sentence
issues in most studies have been analyzed in terms of contractual
structure which contains conditional clauses using the words if
relationships, which existed in the written contract documents;
or unless.
however, the potential issues and their effects caused by the facts
• Step 4. Semantic analysis using contract risk–related lexicon:
not in the written document have not been considered so far.
semantic rules that recognize the pattern of the contractor-
friendly sentences and expressions are developed depending on
Research Approach types of contract risks defined in the first step. The construction
contract risks–related lexicon is applied to the process of seman-
To address the aforementioned knowledge gaps, this study aims to tic rule matching, and every SVO tuple is matched with con-
develop a systematic approach to identify missing contractor- cepts of the lexicon to make an inference about the meaning
friendly contract clauses, called the Identification Model of Missing of each element. Finally, the judgement of whether the input
Contract Clauses (IMMC2 ). Careful investigation based on in-depth sentence is a risk or not is made based on the previous semantic
content analysis of contract conditions is required to determine rule-matching process.

© ASCE 05020003-3 J. Manage. Eng.

J. Manage. Eng., 2020, 36(3): 05020003

 Step 1
Category of contractor’s-friendly clause
Identifying contractor- Contract
Precedents
Assignment DSC Employer’s financial Adjustment for
friendly clause (previous arrangement changes in legislation
Conditions
legal case) Extension of time Settlement of Limitation of
Variation Payment
for completion dispute liability
Collecting 17 contracts of the
FIDIC-based project

Step 2
Simplification Rule Development
Text pre-processing The contractor
IF,
using simplification rules shall be
word1 | sentence1 + " : " + " (a) " + sentence2 + " , " | " ; "+ " (b) " + sentence3 + " , " | " ; " + ...
adjusted ….
THEN,
or in the word1 | sentence1 + sentence2;
judicial … or word | sentence1 + sentence3; …
Downloaded from ascelibrary.org by University College London on 03/01/20. Copyright ASCE. For personal use only; all rights reserved.

official, and
… IF,
<subject 1> + "and" | " , "| "or" + <subject 2> + (<MD>) + <verb 1> + (<object 1>)
THEN,
<subject 1> + (<MD>) + <verb 1> + (<object 1>);
<subject 2> + (<MD>) + <verb 1> + (<object 1>)

Step 3
Subject
Neither party
SVO tuple extraction (S)

Verb
shall assign
(V)
the whole or any part
Object of the Contract or
(O) any benefit or
Syntactic analysis and SVO tuple extraction interest in or under
the Contract.
Step 4
Extracted Element
IF,
Semantic analysis subject (NP) == <both-party> class,
Subject
Neither party Semantic rule verb(NP) == "not" + <assign> class,
(S)
object(NP) == <contract> class | <right> class,
Verb THEN,
shall assign
(V) "This is about prohibition of transferring contract. The statement should
the whole or any part of the be included! "
Object Contract or any benefit or ELSE,
(O) interest in or under the
"Prohibition of transferring contract-related clause is missing!"
Contract.

Fig. 1. Overview of the proposed research.

Step 1: Identifying Contractor-Friendly Clauses The Employer shall submit, before the Commencement Date
and thereafter within 28 days after receiving any request from
A case study is used to demonstrate the proposed framework, and
the Contractor, reasonable evidence that financial arrange-
real-world project cases were analyzed to understand what type
ments have been made and are being maintained which will
of contractor-friendly clauses are omitted in the owner’s modified
enable the Employer to pay the Contract Price punctually (as
contract. First, contractor-friendly clauses that are likely disappeared
estimated at that time) in accordance with Clause 14 [Contract
from the owner’s modified contract conditions were identified, build-
Price and Payment].
ing on 17 contract documents of FIDIC-based international construc-
tion projects in which global construction companies participated in This is an important owner prequalification to guarantee a con-
the bidding. In addition, the authors have examined which standard tractor’s right to be paid because it is a great risk to the contractor if
contract conditions are frequently missing, and how the contractor the owner declares that they are unable to pay the contract price or
could be at risk from these missing conditions by comparing the if bankruptcy occurs unexpectedly. Nonetheless, among the
FIDIC standard conditions with the collected 17 contracts. Three adopted 17 sets of contracts, many deliberate deletions of the clause
contract management experts with more than 20 years of experience related to the financial ability of the owner have been discovered.
in the construction field participated in the analysis of the collected One of the precedent cases introduced by Jervis and Levin (1988)
contracts for identifying contractor-friendly clauses. In addition, dis- showed that the contractor was in great danger as a result of the
pute cases due to the missing contractual clauses were examined, and owner’s bankruptcy during a project where the owner’s financial
contractor-friendly clauses that should not be omitted from the con- arrangements clause was missing.
tract conditions were defined. By comparing the FIDIC standard contract with the collected
The following is an example of a frequent missing case in the contracts and examining related precedents, the contractor-friendly
owner’s modified construction contract. The employer (i.e., owner) clauses were categorized as follows:
should generally prove whether they have the financial ability to • Assignment: a condition that neither party shall assign the whole
pay the contract price in accordance with FIDIC (1999) Chapter or any part of the contract or any benefit or interest in or under
2.4, titled “Employer’s Financial Arrangements”: the contract.

© ASCE 05020003-4 J. Manage. Eng.

J. Manage. Eng., 2020, 36(3): 05020003

 Table 1. Examples of contractor-friendly clauses and expressions in this study
Type Examples of contractor-friendly clauses
Assignment Neither party shall assign the whole or any part of the contract or any benefit or interest in or under the contract.
DSC To the extent which was practicable (taking account of cost and time)a, the contractor shall be deemed to have obtained all
necessary information as to risks, contingencies and other circumstances which may influence or affect the tender or works.
Employer’s financial The Employer shall submit, before the commencement date and thereafter within 28 days after receiving any request from the
arrangement contractor, reasonable evidence that financial arrangements have been made and are being maintained which will enable the
employer to pay the contract price punctually (as estimated at that time).
Adjustment for changes in The contract price shall be adjusted to take account of any increase or decrease in cost resulting from a change in the laws of
legislation the country (including the introduction of new laws and the repeal or modification of existing laws) or in the judicial or official
governmental interpretation of such laws, made after the base date, which affect the contractor in the performance of
obligations under the contract.
Payment If the contractor does not receive payment, the contractor shall be entitled to receive financing charges compounded monthly
Downloaded from ascelibrary.org by University College London on 03/01/20. Copyright ASCE. For personal use only; all rights reserved.

on the amount unpaid during the period of delay.

Extension of time for The contractor shall be entitled to an extension of the time for completion if and to the extent that completion for the purposes
completion of sub-clause 10.1 (taking-over of the works and sections) is or will be delayed.
Settlement of dispute Any dispute between the parties arising out of or in connection with the contract not settled amicably and in respect of which
the DB’s decision (if any) has not become final and binding shall be finally settled by arbitration.
Variation The contractor shall execute and be bound by each variation, unless the contractor promptly gives notice to the engineer
stating (with supporting particulars) that the contractor cannot readily obtain the goods required for the variation.
Limitation of liability Neither party shall be liable to the other party for loss of use of any works, loss of profit, loss of any contract or for any indirect
or consequential loss or damage which may be suffered by the other party in connection with the contract.
a
In the case of this clause, “to the extent which was practicable (taking account of cost and time)” is a contractor-friendly expression that is likely to be absent
from the owner’s modified contracts.

• Differing site conditions (DSC): a condition under which the conjunctions) frequently used in the contract clauses. For this reason,
contractor is responsible for the site conditions only to the extent appropriate data preprocessing that eliminates unnecessary informa-
that is practicable (taking account of cost and time). tion and simplifies the syntax structures is required to facilitate the
• Employer’s financial arrangement: a condition that the employer data analysis of the subsequent process.
shall submit reasonable evidence that financial arrangements The preprocessing involves three steps: (1) sentence seg-
have been made and are being maintained that will enable the mentation, (2) removal of stop words, and (3) simplification rule-
employer to pay the contract price. matching. These steps simplify compound and complex syntax
• Adjustment for changes in legislation: a condition that the con- structure of contract clauses. The input text document (i.e., contract
tract price shall be adjusted to take account of any increase in conditions) is first segmented by the sentence unit, and several stop
cost resulting from a change in the laws of the country. words are removed through preprocessing. This study defined some
• Payment: a condition that if the contractor does not receive pay- terms and expressions including if any and as aforesaid, which are
ment, the contractor shall be entitled to receive financing charges. repeated in the contract documents but are less meaningful in the
• Extension of time for completion: a condition that the contractor sentences, as stop words. In most preprocessing of English text
shall be entitled to extend the time for completion if some ex- data, it is common to define words such as a and the, for example,
ceptionally adverse conditions happen. as stop words. In case of the construction contract, however, the
• Settlement of dispute: a condition that any dispute between the meaning can be changed depending on whether the or any is placed
parties shall be finally settled by arbitration, not the decision of in front of the noun, and this can affect the interpretation of con-
employer’s side. tractual facts. Therefore, rather than the stop words used in more
• Variation: a condition that the contractor can give notice to the general English text analysis, only a few words are defined as stop
engineer stating that the contractor cannot readily obtain the words in the study, considering the fact that reliability of the text
goods required for a variation. analysis likely be lowered when these words are excluded from the
• Limitation of liability: a condition that the total liability of the preprocessing step.
contractor to the employer shall not exceed the accepted con- To simplify syntax structure of segmented sentences, simpli-
tract amount. fication rules have been developed in this study. To develop
These nine types of contractor-friendly clauses are fed into the simplification rules, the authors have firstly collected compound,
development of risk identification rules, which are key elements for complex, and compound-complex sentences from the case projects.
the proposed IMMC2 . Table 1 presents examples of contractor- Complex syntax structures, including multiple dependent clauses
friendly clauses. rather than simple sentence forms, were collected, and the patterns
of complex syntax were analyzed to develop simplification rules.
Simplification rules simplify the sentence structure written in par-
Step 2: Text Preprocessing Using Simplification Rules
allel structures using conjunctions, such as and, or, or a colon,
Based on a thorough review of contractor-friendly clauses in the based on if-then logic. Therefore, the simplified sentence structure
previous section, the authors have designed a text data-driven risk thus has the effect of improving the accuracy of the SVO tuple def-
identification model, IMMC2 . Most of the sentences included in a inition process for subsequent analysis.
construction contract have complex structures, including a plurality
of subordinate clauses rather than a simple sentence form (Lee et al.
Step 3: SVO Tuple Extraction
2019). Thus, one of the challenges of analyzing contract documents
is due to many specialized vocabularies and complex syntactic struc- Syntactic analysis in this study involves analyzing the grammatical
tures of sentences (e.g., parallel structures with many coordinating structure of components in a sentence and assigning structured

© ASCE 05020003-5 J. Manage. Eng.

J. Manage. Eng., 2020, 36(3): 05020003

 Class A : Environment Class B : Resource Class C : Actor

Legal environment Sub-class1 Work Construction Party Employer

Sub-class1 Sub-class1 Employer-side
Sub-class2 Goods
Sub-class2 Legal Party Contractor-side Client
Sub-class2 Physical environment
Sub-class3 Document
Subcontractor-side Manager
Sub-class3 Third Party
Sub-class4 Information (Notice) Engineer
Sub-class3 Force Majeure
Time Sub-class4 Both Party
Sub-class5 …

Class D : Payment Class E : Action Class F : Process Class G : Right &

Responsibility
Sub-class1 Currency Sub-class1 Right
Sub-class1 Administrative action Sub-class1 Construction process
Downloaded from ascelibrary.org by University College London on 03/01/20. Copyright ASCE. For personal use only; all rights reserved.

Sub-class2 Payment amount Sub-class2 Liability

Sub-class3 Payment certificate Sub-class2 Unfavorable action Sub-class2 Legal process Sub-class3 Priority

Sub-class4 Cost Sub-class4 Level of right / liability

Fig. 2. Contract risk–related lexicon taxonomy.

sentence components to SVO tuple. That is, unstructured text data To develop a contract risk–related lexicon, the authors have
are converted into structured data by extracting SVO tuples to auto- collected basic and comprehensive English vocabularies from
matically understand the pattern of the sentence in the contract WordNet (Fellbaum 2012), which is not domain-dependent, and
document. All segmented and simplified sentences from the pre- is a popular lexical dictionary applied to many NLP researches
vious steps go through syntactic analysis including determining (Kamps et al. 2004; Shehata 2009; Zhang and El-Gohary 2016)
parts of speech and sentence structure. In this step, part of speech for semantic analysis. Because vocabularies from WordNet are
(POS) tagging and dependency grammar are used to analyze the not domain-dependent, however, additional definition of some
structure of the sentence based on the dominance-dependence re- words that are differently used in the construction contract domain
lationship between words. is required. For example, engineer is defined in WordNet as a person
In the process of syntactic analysis of contract documents, syn- who uses scientific knowledge to solve practical problems; however,
tactic rules were developed in order to extract the SVO information engineer is typically used in the construction contract domain as
for an exceptional sentence structure that contains conditional a contract manager or project manager, for example. Therefore,
clauses using if or unless. Syntactic rules help to distinguish SVO such domain-dependent vocabularies were collected building upon
tuple of subordinate clause from the SVO tuple of main clause in a previous studies related to contract vocabulary (Kabilan and
sentence. If an input sentence contains conditional clauses using Johannesson 2003; Al Qady and Kandil 2010; Yan et al. 2006).
unless, the SVO tuple should be separately extracted from the un- Fig. 2 is a taxonomy of the contract risk–related lexicon defined
less clause and the main clause, respectively, to understand the in this study showing the hierarchy of words and expressions. All of
semantic dependent relationship among clauses. That is, the con- the vocabularies in the lexicon can be separated into unigram, bi-
tractual action of the unless clause determines and subordinates the grams, and tirgrams, and a total of 352 vocabularies were collected.
contractual action of the main clause. Therefore, the respective The contract risk–related lexicon built through this process is used
SVO tuple of each clause is required, even if it is one single sen- in the semantic analysis stage of the IMMC2 , which allow one to
tence. In other words, syntactic rules are needed in order to extract
infer the background knowledge and semantic relations of the sen-
the SVO tuple by separating the if/unless clause and the main
tence related to the contractor-friendly clauses.
clause to better understand the semantic relationship among the
IMMC2 learns patterns of contractor-friendly clauses defined
sentence.
and recognizes sentences similar to the learned patterns in order
to detect whether the contractor-friendly clauses are missing or
Step 4: Semantic Analysis Using Contract not in given contracts. For this purpose, it is necessary to develop
Risk–Related Lexicon semantic rules that recognize the pattern of the contractor-friendly
For the semantic analysis of sentences, understanding similar ex- sentences and expressions. In other words, if the tuple of a sentence
pressions such as synonyms and the hierarchy of words such as of a contract is similar to the contractor-friendly clauses, it can be
hypernyms and hyponyms are important. In the process of NLP, judged that the contract contains the contractor-friendly clauses.
a lexicon is used for inferring the semantic relations between words However, if none of the sentences are identical with the SVO pat-
in the sentences. Not only the unique information (i.e., part of terns of semantic rules, the contract will be deemed to be missing
speech and grammatical form) of the words in the sentence, but the contractor-friendly clauses, indicating that the contractor has
also the information on the type of semantic class to which the word potential risk. The SVO patterns for the contractor-friendly clauses
belongs and the type of hypernym that exists could be understood were defined with reference to the class and concepts of the con-
through the lexicon (Lee et al. 2019). Thus, a lexicon specialized in tract risks–related lexicon.
the construction contract domain is an important clue for under- To analyze the semantic relations between the elements of the
standing semantic relations of sentences and matching semantic sentence, the SVO tuple information extracted from the syntactic
rules for judging whether the sentence is a contractor-friendly analysis step is compared with the SVO tuple of the contractor-
or not. friendly clauses collected. Fig. 3 illustrates the rule-matching

© ASCE 05020003-6 J. Manage. Eng.

J. Manage. Eng., 2020, 36(3): 05020003

 Semantic Rules Contract Document
(SVO pattern) (SVO pattern)

SR 1 S1
if, SR1 = S1 or S2 or S3 … Sk
SR 2 S2 then, a contractor-friendly clause (SR1) is included!
SR 3 S3
. .
. . if, SRn S1 or S2 or S3 … Sk
. .
then, contractor-friendly clauses are missing!
SR n Sk

Fig. 3. Rule-matching structure for semantic rules and contract sentences.

Downloaded from ascelibrary.org by University College London on 03/01/20. Copyright ASCE. For personal use only; all rights reserved.

process for semantic rules and contract sentences. Semantic rules imported document, and the model is terminated when analysis
based on contractor-friendly clauses are defined as pattern of SVO of the last sentence of the document is complete.
tuple from semantic rules (SR) SR1 to SRn according to the nine Through case studies, the authors evaluated how accurately the
risk types of clause. In addition, the input contract document is seg- model can identify missing clauses in the owner’s modified con-
mented into a sentence unit (i.e., S1 to Sk ) through preprocessing, tract conditions. Tables 2 and 3 present examples of whether the
and the individual sentences of the contract document are also con- proposed model can correctly identify defined missing clauses
verted into the SVO tuple type through a syntactic analysis step. in the contract document of the case project. For example, the first
Therefore, the SVO tuples of semantic rules and the SVO tuples missing clause in Table 2 is an example of missing the payment risk
of a contract document can be examined one by one. If any of type clause that occurred because a certain semantic rule of a pay-
the patterns of the sentences in the contract (Sk ) match the semantic ment risk type did not match any of the SVO tuples of the modified
rules (SRn ), then the contractor-friendly clause of the risk type cor- contract document. By doing so, the model informs users (i.e., con-
responding to the semantic rule is deemed to be included in the tractor) of what type of risk is missing in the contract, supporting
contract. However, if a certain type of semantic rule does not match the users to develop an appropriate response strategy for the given
any of the sentences in the document, it means that the required risk.
contractor-friendly clause of certain risk type is not included
in the given contract. In this case, the certain risk type of a
contractor-friendly clause is deemed to be omitted from the contract Evaluation and Discussion
document.
To evaluate the reliability of proposed model, the performance
of the proposed model was compared against human tasks. Five
Model Implementation of the Proposed Approach experts with 5–20 years of experience in the field of construction
contract review were selected using the snowball sampling method,
Based on the previous step-by-step process and related rules, and they participated in the contract review process in a nonface-
IMMC2 was implemented for testing. The developed model has to-face manner. The results of the contract reviews by the domain
three different modules (i.e., preprocessing, syntactic analysis, and experts was received via email after instructing the purpose of this
semantic analysis module) as shown in Fig. 4. evaluation in advance. The authors then compared the results of the
The preprocessing module restructures the unstructured input contract review by the experts with the results of the contract review
data so that they can be understood by the computer, and also ex- by IMMC2 in order to evaluate whether the missing clauses were
cludes unnecessary information. In the preprocessing phase, which correctly detected. One of the case project’s contract conditions,
is the first stage of IMMC2 , the input data of the text document are which include some of omitted contractor-friendly clauses, was
separated into individual sentences, and the sentence simplification adopted for the evaluation. The applied case project is an office
is performed using simplification rules. The simplified sentence is building construction project, and the construction period is about
processed through the syntactic analysis including parts of speech 30 months. The human experts and the proposed model reviewed
of each word and analysis of sentence structure. the same contract conditions.
After the syntactic analysis, the syntactic rule-matching process The performance of proposed model was evaluated using preci-
is performed to correctly extract SVO tuple of the sentence. In the sion, recall, and F-measure. Precision measures the percentage of
semantic analysis module, the model determines whether the pat- missing clauses correctly identified from the total number of clauses
terns of the subject, verb, and object tuples in the contract condi- identified, and recall measures the percentage of missing clauses cor-
tions are similar to the patterns of the predefined contractor-friendly rectly identified from the total number of missing clauses existing in
clauses. The contract risks–related lexicon is applied to the pro- the source text. When compared with human evaluators, IMMC2
cess, and the semantic relations and combination of each element achieved 85.7% precision, 75% recall, and an F-measure of 80%.
are finally judged in terms of whether or not that element is re- As a result of the contract review of human experts, it was found
lated to certain type of risk (i.e., whether or not a certain type of that the results differ depending on individual knowledge and ex-
contractor-friendly clause is missing). If the pattern of imported perience of each expert. In particular, some missing contractor-
SVO tuple matches a certain predefined pattern, then the contract friendly clauses were not mistakenly recognized by human experts.
document will be deemed to contain the corresponding contractor- One of experts found every missing contractor-friendly clause in
friendly clause in the contract. Otherwise, a warning message will the contract, but one of them found only some of such sentences.
be displayed to help users recognize such a problem. The entire In other words, the human tasks show that manually identifying
process of the model is iterative for every single sentence in the missing clauses in a contract review process is labor-intensive,

© ASCE 05020003-7 J. Manage. Eng.

J. Manage. Eng., 2020, 36(3): 05020003

 Contract Conditions Pre-processing Module

S1. Engineer will

decide all questions … ‘Engineer’ ‘will’ ‘decide’ ‘all’
S2. … ‘questions’ …
S3. … .
. Syntactic analysis Module
.
Sk. … Simplification Syntactic
Rules Rules

POS tag set

(‘Engineer’, ‘NN’), (‘will’, ‘MD’), NN : Noun, singular or mass
Downloaded from ascelibrary.org by University College London on 03/01/20. Copyright ASCE. For personal use only; all rights reserved.

(‘decide’, ‘VB’), (‘all’, ‘DT’), MD : Modal

(‘questions’, ‘NNS’)… VB : Verb, base form
DT : Determiner
NNS : Noun, plural

Subject Verb Object

Part Part Part

Semantic analysis Module

Semantic Rules DB

Subject Verb Object Class 1

Part Part Part Class 2 Contract Risk
Class 3 Lexicon
…

Pattern of contractor-friendly
clause

(Class 1, Class 2, Class 3)

Contractor-friendly clause!

Fig. 4. Operation process of IMMC2 .

time-consuming, and error-prone, and the proposed model thus has Second, there was an error that occurred when a sentence that
the potential to enhance the process of reviewing for omitted does not correspond to the contractor-friendly clauses is errone-
contractor-friendly clauses. Moreover, human experts took 1–2 days ously identified as a contractor-friendly sentence. This is a case
to review the whole contract conditions; however, the computing when a pattern of irrelevant sentence is similar to the pattern de-
time of IMMC2 was about 2–3 min on average with a 1.80 GHz, fined in the semantic rules; thus, the irrelevant sentence is mistak-
Intel Core processor and 16 GB RAM. enly identified as a contractor-friendly clause. Such an error is
Some error types of the implemented model, which have the increased when the semantic rules are defined as too general pat-
potential to improve the proposed model, were found in the evalu- terns. Therefore, it is necessary to define the semantic rules more
ation process as follows. First, there was an error that occurred specifically so that irrelevant sentences are not identified.
when a pattern of an imported contractor-friendly clause was com- Looking at the two aforementioned errors, it seems that there is
pletely different from the learned patterns of contractor-friendly a trade-off relationship between them. In other words, if the range
clauses used in the process of developing the semantic rules. As of the sentences that match the rules is narrowed excessively by
a result, the semantic rule matching was not properly performed, defining the semantic rules too specifically to reduce the second
and thus the model concluded that a contractor-friendly clause was type of errors, it becomes possible to identify only the sentences
missing even though the contractor-friendly clause existed in the of the patterns completely matching the semantic rules. This can
contract. This type of error is a typical error that occurs in most lead to errors that do not properly identify similar patterns or
rule-based information extraction models, and the error occurs not new patterns of sentences in the contract. Conversely, if the match-
only because of the infinite amount of data that can be used to de- ing scope of the rules is too far broad, there exists a risk that the
velop the rules, but also because of the limitation whereby it is not model identifies unrelated sentences, which is problematic. As a
possible to develop every pattern of data as rules (Lee et al. 2019). result, if the rule is specified to reduce the second type error, the
To reduce such errors, defining generalized semantic rules that can- first type error is more likely to occur, and if the rule is generalized
not miss the similar patterns is necessary when adding or modifying to reduce the first type error, the possibility of the second type error
semantic rules. increase vice versa. Therefore, it is necessary to prioritize the types

© ASCE 05020003-8 J. Manage. Eng.

J. Manage. Eng., 2020, 36(3): 05020003

 Downloaded from ascelibrary.org by University College London on 03/01/20. Copyright ASCE. For personal use only; all rights reserved.

Table 2. Examples of results of model application

Risk type

© ASCE
No. Missing clauses SVO SVO tuple A B C D E F G H I Final message
1 The contractor shall be S The contractor shall be — — — — X — — — — Contractor’s right to receive
entitled to receive financing V entitled to receive financing financing charges caused by
charges compounded O charges compounded delayed payment-related
monthly on the amount monthly on the amount clause is missing!
unpaid during the period of unpaid during the period of
delay. delay.
2 The contractor shall execute (Main clause) S The contractor shall execute — — — — — — — X — Variation-related contractor-
and be bound by each (Main clause) V and be bound by each friendly clause is missing!
variation, unless the (Main clause) O variation the contractor
contractor promptly gives (Unless clause) S promptly gives notice to the
notice to the engineer stating (Unless clause) V engineer stating (with
(with supporting particulars) (Unless clause) O supporting particulars) that
that the contractor cannot the contractor cannot readily
readily obtain the goods obtain the goods required
required for the variation. for the variation.
Note: Risk type: A = assignment; B = DSC; C = employer’s financial arrangement; D = adjustment for changes in legislation; E = payment; F = extension of time for completion; G = settlement of dispute; and
H = variation.

05020003-9
Table 3. Further examples of results of model application
Risk type
No. Missing clauses SVO SVO tuple A B C D E F G H I Final message

J. Manage. Eng., 2020, 36(3): 05020003

3 Any dispute between the parties arising out of or in S Any dispute between the parties arising out of or in — — — — — — X — — Error
connection with the contract not settled amicably V connection with the contract not settled amicably and
and in respect of which the DB’s decision (if any) O in respect of which the DB’s decision has not become
has not become final and binding shall be finally final and binding shall be finally settled by arbitration
settled by arbitration.
4 The contract price shall be adjusted to take account S The contract price shall be adjusted to take account of — — — X — — — — — This statement is
of any increase or decrease in cost resulting from a V any increase or decrease in cost resulting from about adjustments
change in the laws of the country or in the judicial O a change in the laws of the country or in the judicial for changes in
or official governmental interpretation of such laws, or official governmental interpretation of such laws, legislation. The
made after the base date, which affect the contractor made after the base date, which affect the contractor statement should be
in the performance of obligations under the in the performance of obligations under the contract. included!
contract.
5 The total liability of the contractor to the employer S The total liability of the contractor to the employer — — — — — — — — X Limitation of
shall not exceed the sum resulting from the V shall not exceed the sum resulting from the liability-related
application of a multiplier (less or greater than one) O application of a multiplier (less or greater than one) contractor-friendly
to the accepted contract amount, as stated in the to the accepted contract amount, as stated in the clause is missing!
contract data, or (if such multiplier or other sum is contract data, or (if such multiplier or other sum is not
not so stated), the accepted contract amount. so stated), the accepted contract amount.

Note: Risk type: A = assignment; B = DSC; C = employer’s financial arrangement; D = adjustment for changes in legislation; E = payment; F = extension of time for completion; G = settlement of dispute; and

J. Manage. Eng.
H = variation.
 of errors depending on the error’s impact on the performance and reason why this study applied the rule-based NLP to develop iden-
usability of the model. tification rules. Therefore, the developed identification rules have
In the case of applying NLP for the purpose of reviewing con- limitations that hardly cover each and every possible modified con-
tract risks in this study, the second type error involves more risk tract condition. The proposed model has the potential for continuous
than the first type error in terms of future impact. This is because updating of the identification rules, and if a large amount of contract
misidentifying a missing sentence as an unmissed sentence is more data can be obtained for the future works, further improved perfor-
critical than misidentifying an unmissed sentence as a missing sen- mance is expected through a machine learning–based approach.
tence. Therefore, this study sought to further reduce the second type
error, which recognizes that a contractor-friendly clause is not miss-
ing even though a contractor-friendly clause is missing in the Data Availability Statement
contract.
Some or all data, models, or code generated or used during the
study are available from the corresponding author by request, such
Downloaded from ascelibrary.org by University College London on 03/01/20. Copyright ASCE. For personal use only; all rights reserved.

Summary and Conclusion as lexicon and identification rules.

This study has offered a domain-specific new model-based ap-

proach for identifying missing contractor-friendly clauses from the Acknowledgments
owner’s modified contract. The case studies demonstrated that the
developed model could automatically identify missing contractor- This research was supported by Basic Science Research Program
friendly clauses in the owner’s modified contract based on FIDIC through the National Research Foundation of Korea (NRF) funded
standard conditions. Rule-based NLP was applied in the case stud- by the Ministry of Education (NRF-2018R1D1A1A09083708 and
ies to develop a new proactive risk assessment model. The model, NRF-2018R1A6A3A01012406).
IMMC2 , identifies the contractor-friendly clauses that are omitted
from the owner’s modified contract conditions and help contractors References
to double-check the key clauses that should not be missed at the
bidding and contracting stage. The IMMC2 contains three main Abotaleb, I. S., I. H. El-adaway, and M. B. Moussa. 2019. “Guidelines for
modules: (1) preprocessing, (2) syntactic analysis, and (3) semantic administrating and drafting nonpayment owners’ obligation provisions
analysis. Semantic rules have been developed in order to examine under design-build contracts.” J. Manage. Eng. 35 (4): 04019010.
the omission of contractor-friendly clauses in the contract, and the https://doi.org/10.1061/(ASCE)ME.1943-5479.0000693.
model has been applied to real-world case projects. Al Qady, M., and A. Kandil. 2010. “Concept relation extraction from con-
struction documents using natural language processing.” J. Constr. Eng.
The contributions of this study can be summarized as follows.
Manage. 136 (3): 294–302. https://doi.org/10.1061/(ASCE)CO.1943
First, the proposed model can support the process of reviewing con- -7862.0000131.
tract conditions from the contractor’s point of view, which helps Arcadis. 2018. Global construction disputes report 2018: Does the con-
alleviate potential human mistakes by automatically finding miss- struction industry learn from its mistakes? Amsterdam, Netherlands:
ing contractor-friendly clauses. Given that finding missing clauses Arcadis.
is more likely to be subject to a mistake than finding problematic ASCE Insurance in Contracts. 2015. What you need to know about liability
clause by human works, this research work is expected to be effec- clauses—Risk management resources. Peoria Heights, IL: ASCE Group
tively used for minimizing human errors and supporting contract Insurance Program Administrator.
management tasks. Second, this study offers a domain-specific Awwad, R., B. Barakat, and C. Menassa. 2016. “Understanding dispute
resolution in the Middle East region from perspectives of different
lexicon that can assist in capturing domain-specific semantic con-
stakeholders.” J. Manage. Eng. 32 (6): 05016019. https://doi.org/10
cepts in the construction contract documents. Third, this study ul- .1061/(ASCE)ME.1943-5479.0000465.
timately supports proactive risk management in the bidding and Bubshait, A. A., and S. A. Almohawis. 1994. “Evaluating the general
contracting phase of the construction project by developing a quick conditions of a construction contract.” Int. J. Project Manage. 12 (3):
and robust decision support tool for contractors. 133–136. https://doi.org/10.1016/0263-7863(94)90027-2.
This research is expected to provide a data-driven contract man- Butcher, T., S. Newboult, and H. Rahman. 2018. “FIDIC 2017: Risk allo-
agement approach that helps to understand the contract risk hidden cation under the updated forms of contract, construction guides series.”
behind the contract documents. The proposed model takes the proj- DLA Piper. Accessed September 22, 2019. https://www.dlapiper.com
ect’s contract document as input data and automatically reviews the /en/abudhabi/insights/publications/2018/03/constructive-thinking-roundup
missing contractor-friendly clauses in an automated manner. How- -february-2018/fidic-2017-risk-allocation/.
Cakmak, P. I., and E. Cakmak. 2013. “An analysis of causes of disputes in
ever, the proposed model does not intend to completely replace the
the construction industry using analytical hierarchy process (AHP).” In
human work for contract review. Rather, the model aims to comple- Proc., Architectural Engineering Conf., 94–102. Reston, VA: ASCE.
ment such human work as a support tool to double-check the con- Chen, C., C. H. Jiang, and H. J. Li. 2013. “Modification to the FIDIC silver
tract conditions. book in a buyers market context: A case study of the Hai Phong power
The findings of this study need to be interpreted in the light of plant II project.” In Applied mechanics and materials, 357–360. Zurich,
the following limitations. First, because the IMMC2 is based upon Switzerland: Trans Tech.
the FIDC standard contract conditions, the model has limitations in Cheung, S.-O. 1999. “Critical factors affecting the use of alternative dispute
generality when analyzing other types of contracts such as JCT and resolution processes in construction.” Int. J. Project Manage. 17 (3):
AIA standard contracts. Therefore, it is expected that future studies 189–194. https://doi.org/10.1016/S0263-7863(98)00027-1.
Chong, H.-Y., and R. M. Zin. 2010. “A case study into the language struc-
will expand the usability of the developed model by collecting pat-
ture of construction standard form in Malaysia.” Int. J. Project Manage.
terns and learning patterns from vocabularies and sentences used in 28 (6): 601–608. https://doi.org/10.1016/j.ijproman.2009.09.008.
other types of standard contract conditions. Second, collected data Craig, D. 2016. “Top 4 construction contract provisions to be negotiated
were relatively small amounts compared with huge amounts of big with the owner.” Accessed September 22, 2019. https://jobsite.procore
data to build learning data on patterns of contractor-friendly clauses .com/top-4-construction-contract-provisions-to-be-negotiated-with-the
because contract documents are not easy to access. That was the -owner/.

© ASCE 05020003-10 J. Manage. Eng.

J. Manage. Eng., 2020, 36(3): 05020003

 Eggleston, B. 2009. Liquidated damages and extensions of time: In con- Marzouk, M., and M. Enaba. 2019. “Text analytics to analyze and monitor
struction contracts. New York: Wiley. construction project contract and correspondence.” Autom. Constr.
Fellbaum, C. 2012. “WordNet.” The encyclopedia of applied linguistics. 98 (Feb): 265–274. https://doi.org/10.1016/j.autcon.2018.11.018.
New York: Wiley. Milner, A. 2011. “Contract interpretation: Potential for relaxing the exclu-
FIDIC (International Federation of Consulting Engineers). 1999. Condi- sionary rule.” Int. J. Law Built Environ. 3 (3): 205–221. https://doi.org
tions of contract for construction for building and engineering works /10.1108/17561451111178425.
designed by the employer. Geneva: FIDIC.
Nguyen, D. A., M. J. Garvin, and E. E. Gonzalez. 2018. “Risk allocation in
Gao, N., Y. Chen, W. Wang, and Y. Wang. 2018. “Addressing project com-
U.S. public-private partnership highway project contracts.” J. Constr.
plexity: The role of contractual functions.” J. Manage. Eng. 34 (3):
04018011. https://doi.org/10.1061/(ASCE)ME.1943-5479.0000613. Eng. Manage. 144 (5): 04018017. https://doi.org/10.1061/(ASCE)CO
Hamie, J., and M.-A. Abdul-Malak. 2018a. “Rules-based approach for .1943-7862.0001465.
construction contract documents interpretation.” In Proc., Construction Niu, J., R. R. A. Issa, and I. Mutis. 2015. “Taxonomy development toward
Research Congress 2018, 186–195. Reston, VA: ASCE. the domain ontology of construction contracts: A case study on AIA
Hamie, J. M. K., and M.-A. U. Abdul-Malak. 2018b. “Model language A201-2007.” In Ontology in the AEC industry, 217–250. Reston,
Downloaded from ascelibrary.org by University College London on 03/01/20. Copyright ASCE. For personal use only; all rights reserved.

for specifying the construction contract’s order-of-precedence clause.” VA: ASCE.

J. Leg. Aff. Dispute Resolut. Eng. Constr. 10 (3): 04518011. https://doi On Cheung, S., L. Zhu, and K. Wai Lee. 2018. “Incentivization and inter-
.org/10.1061/(ASCE)LA.1943-4170.0000260. dependency in construction contracting.” J. Manage. Eng. 34 (3):
Hughes, W., R. Champion, and J. Murdoch. 2015. Construction contracts: 04018010. https://doi.org/10.1061/(ASCE)ME.1943-5479.0000601.
Law and management. Abingdon, UK: Routledge. Rameezdeen, R., and A. Rodrigo. 2014. “Modifications to standard forms
Jervis, B. M., and P. Levin. 1988. Construction law, principles and prac- of contract: The impact on readability.” Australas. J. Constr. Econ.
tice. New York: McGraw-Hill College. Build. 14 (2): 31. https://doi.org/10.5130/AJCEB.v14i2.3778.
Kabilan, V., and P. Johannesson. 2003. “Semantic representation of contract
Semple, C., F. T. Hartman, and G. Jergeas. 1994. “Construction claims and
knowledge using multi tier ontology.” In Proc., 1st Int. Conf. on Seman-
disputes: Causes and cost/time overruns.” J. Constr. Eng. Manage.
tic Web and Databases, 378–397. CEUR-WS.org.
120 (4): 785–795. https://doi.org/10.1061/(ASCE)0733-9364(1994)
Kamps, J., M. Marx, R. J. Mokken, and M. De Rijke. 2004. “Using Word-
Net to measure semantic orientations of adjectives.” In Vol. of 4 Proc., 120:4(785).
LREC-04, 4th Int. Conf. on Language Resources and Evaluation, Sertyesilisik, B. 2010. “Investigation on particular contractual issues in
1115–1118. Lisbon, PT. construction.” J. Leg. Aff. Dispute Resolut. Eng. Constr. 2 (4): 218–227.
Laryea, S. 2011. “Quality of tender documents: Case studies from the UK.” https://doi.org/10.1061/(ASCE)LA.1943-4170.0000032.
Constr. Manage. Econ. 29 (3): 275–286. https://doi.org/10.1080/01446193 Shehata, S. 2009. “A WordNet-based semantic model for enhancing text
.2010.540019. clustering.” In Proc., 2009 IEEE Int. Conf. on Data Mining Workshops,
Lee, J. 2018. Development of construction risk extraction model for over- 477–482. New York: IEEE.
seas construction projects based on natural language processing Walsh, K. P. 2017. “Identifying and mitigating the risks created by prob-
(NLP). Seoul: Ewha Womans Univ. lematic clauses in construction contracts.” J. Leg. Aff. Dispute Resolut.
Lee, J., J.-S. Yi, and J. Son. 2019. “Development of automatic-extraction Eng. Constr. 9 (3): 03717001. https://doi.org/10.1061/(ASCE)LA.1943
model of poisonous clauses in international construction contracts using -4170.0000225.
rule-based NLP.” J. Comput. Civ. Eng. 33 (3): 04019003. https://doi.org Yan, Y., J. Zhang, and M. Yan. 2006. “Ontology modeling for contract:
/10.1061/(ASCE)CP.1943-5487.0000807. Using OWL to express semantic relations.” In Proc., IEEE Int. Enter-
Li, K., and S. On Cheung. 2018. “Bias measurement scale for repeated
prise Distributed Object Computing Workshop. New York: IEEE.
dispute evaluations.” J. Manage. Eng. 34 (4): 04018016. https://doi
Youssef, A., H. Osman, M. Georgy, and N. Yehia. 2018. “Semantic risk
.org/10.1061/(ASCE)ME.1943-5479.0000617.
assessment for ad hoc and amended standard forms of construction con-
Lu, W., L. Zhang, and L. Zhang. 2016. “Effect of contract completeness on
contractors’ opportunistic behavior and the moderating role of inter- tracts.” J. Leg. Aff. Dispute Resolut. Eng. Constr. 10 (2): 04518002.
dependence.” J. Constr. Eng. Manage. 142 (6): 04016004. https://doi https://doi.org/10.1061/(ASCE)LA.1943-4170.0000253.
.org/10.1061/(ASCE)CO.1943-7862.0001110. Zaghloul, R., and F. Hartman. 2003. “Construction contracts: The cost of
Maemura, Y., E. Kim, and K. Ozawa. 2018. “Root causes of recurring con- mistrust.” Int. J. Project Manage. 21 (6): 419–424. https://doi.org/10
tractual conflicts in international construction projects: Five case studies .1016/S0263-7863(02)00082-0.
from Vietnam.” J. Constr. Eng. Manage. 144 (8): 05018008. https://doi Zhang, J., and N. M. El-Gohary. 2016. “Extending building information
.org/10.1061/(ASCE)CO.1943-7862.0001523. models semiautomatically using semantic natural language processing
Manning, C. D., C. D. Manning, and H. Schütze. 1999. Foundations of techniques.” J. Comput. Civ. Eng. 30 (5): C4016004. https://doi.org/10
statistical natural language processing. Cambridge, MA: MIT Press. .1061/(ASCE)CP.1943-5487.0000536.

© ASCE 05020003-11 J. Manage. Eng.

J. Manage. Eng., 2020, 36(3): 05020003