Gilles Moyse

Large language models (LLMs) have been shown to be able to perform new tasks based on a few demonstrations or natural language instructions. While these capabilities have led to widespread adoption, most LLMs are developed by resource-rich organizations and are frequently kept from the public. As a step towards democratizing this powerful technology, we present BLOOM, a 176B-parameter open-access language model designed and built thanks to a collaboration of hundreds of researchers. BLOOM is a… Voir plus

Large language models (LLMs) have been shown to be able to perform new tasks based on a few demonstrations or natural language instructions. While these capabilities have led to widespread adoption, most LLMs are developed by resource-rich organizations and are frequently kept from the public. As a step towards democratizing this powerful technology, we present BLOOM, a 176B-parameter open-access language model designed and built thanks to a collaboration of hundreds of researchers. BLOOM is a decoder-only Transformer language model that was trained on the ROOTS corpus, a dataset comprising hundreds of sources in 46 natural and 13 programming languages (59 in total). We find that BLOOM achieves competitive performance on a wide variety of benchmarks, with stronger results after undergoing multitask prompted finetuning. To facilitate future research and applications using LLMs, we publicly release our models and code under the Responsible AI License. Voir moins

Motivated by the lack of data for non-English languages, in particular for the evaluation of downstream tasks such as Question Answering, we present a participatory effort to collect a native French Question Answering Dataset. Furthermore, we describe and publicly release the annotation tool developed for our collection effort, along with the data obtained and preliminary baselines.

This paper investigates the question of the interpretability of fuzzy linguistic summaries, both at the sentence level and at the summary level, seen as a set of sentences. The individual sentence interpretability is examined as depending both on its representativity measured by a quality degree and on its linguistic expression. Different properties at the summary level are also discussed, namely their consistency, their non-redundancy and the information they convey. (C) 2014 Published by… Voir plus

This paper investigates the question of the interpretability of fuzzy linguistic summaries, both at the sentence level and at the summary level, seen as a set of sentences. The individual sentence interpretability is examined as depending both on its representativity measured by a quality degree and on its linguistic expression. Different properties at the summary level are also discussed, namely their consistency, their non-redundancy and the information they convey. (C) 2014 Published by Elsevier B.V. Voir moins

This paper proposes a method to linguistically summarise the local periodic components of a time series: it identifies subparts of the data which are periodic, together with their periodicity degree and period, and provides a linguistic description thereof. The generated sentences can be illustrated by the example " Approximately from March to June, the series is highly periodic with a period of exactly 2 weeks ". The method proposed to identify local periodic zones relies on the determination… Voir plus

This paper proposes a method to linguistically summarise the local periodic components of a time series: it identifies subparts of the data which are periodic, together with their periodicity degree and period, and provides a linguistic description thereof. The generated sentences can be illustrated by the example " Approximately from March to June, the series is highly periodic with a period of exactly 2 weeks ". The method proposed to identify local periodic zones relies on the determination of relevant auto-adaptive windows, based on an analytical expression of the probability distribution of the considered periodicity criterion. The linguistic description generation, in the protoform approach framework, expresses three core aspects of the identified periodic intervals, namely their time context or localisation in time, their periodicity and their period. Intensive experiments performed on both artificial and real data validate the proposed method. Voir moins

An important aspect of interpretability in Fuzzy Linguistic Summaries (FLS) is the absence of opposition therein, which is not guaranteed by the the current approaches used for their generation, possibly leading to confusion for the end-user. In this paper, we first introduce a 3-level hierarchy to organise the models of opposition starting from simpler sentences, then enriched with generalised quantifiers and thirdly considering the several negation operators allowed by fuzzy logic. We then… Voir plus

An important aspect of interpretability in Fuzzy Linguistic Summaries (FLS) is the absence of opposition therein, which is not guaranteed by the the current approaches used for their generation, possibly leading to confusion for the end-user. In this paper, we first introduce a 3-level hierarchy to organise the models of opposition starting from simpler sentences, then enriched with generalised quantifiers and thirdly considering the several negation operators allowed by fuzzy logic. We then introduce a general model of opposition for FLS sentences, which we propose to represent as a 4-dimensional cube. We additionally discuss the antonym property in this analysis framework and prove it for general protoforms. Voir moins

We propose a three-part study of the causes of the complexity to identify contradictory sentences in fuzzy linguistic summaries. The first part deals with the definition of opposition, based on the modern and Aristotelian squares. The second one covers the generalized quantifiers, richer and more complex than « All » and « Some ». Finally, the third one details the properties of the fuzzy logic tools used to model opposition. At the end of our analysis, we propose some ideas to define new… Voir plus

We propose a three-part study of the causes of the complexity to identify contradictory sentences in fuzzy linguistic summaries. The first part deals with the definition of opposition, based on the modern and Aristotelian squares. The second one covers the generalized quantifiers, richer and more complex than « All » and « Some ». Finally, the third one details the properties of the fuzzy logic tools used to model opposition. At the end of our analysis, we propose some ideas to define new properties ensuring non contradiction in the summaries. Voir moins

The erosion score is a Mathematical Morphology tool used primarily to detect periodicity in data. In this paper, three new computation methods are proposed, to decrease its computational cost and to allow to process data streams, in an incremental variant. Experimental results show the signifcant computation time decrease, especially for the efficient levelwise incremental approach which is able to process a one million point data stream in 1.5s.

The paper presents a methodology to evaluate the periodicity of a temporal data series, neither relying on assumption about the series form nor requiring expert knowledge to set parameters. It exploits tools from mathematical morphology to compute a periodicity degree and a candidate period, as well as the fuzzy set theory to generate a natural language sentence, improving the result interpretability. Experiments on both artificial and real data illustrate the relevance of the proposed approach.

Linguistic summarization is a data mining and knowledge discovery approach to extract patterns and sum up large volume of data into simple sentences. There is a large research in generating linguistic summaries which can be used to better understand and communicate about patterns, evolution and long trends in numerical, time series or labelled data. The objective of this work is to develop a computational system capable of automatically generate linguistic descriptions in time series data… Voir plus

Linguistic summarization is a data mining and knowledge discovery approach to extract patterns and sum up large volume of data into simple sentences. There is a large research in generating linguistic summaries which can be used to better understand and communicate about patterns, evolution and long trends in numerical, time series or labelled data. The objective of this work is to develop a computational system capable of automatically generate linguistic descriptions in time series data containing labelled data, not only of the whole series, but also on the differences between subsets of the data. For this purpose we propose a new type of differential summaries, based on a numerical criterion assessing the behaviour of the summary on each subset of interest. Furthermore, this paper proposes an extension of linguistic summaries to provide temporal and categorical contextualisation. This is of particular interest in healthcare to detect differences related to a condition or illness as well as the effectiveness of the administered treatment. Voir moins

The erosion score is a Mathematical Morphology tool used primarily to detect periodicity in data. In this paper, three new computation methods are proposed, to decrease its computational cost and to allow to process data streams, in an incremental variant. Experimental results show the signifcant computation time decrease, especially for the efficient levelwise incremental approach which is able to process a one million point data stream in 1.5s.

Along with the increase of the amount of data stored and to be analyzed, different techniques of data analysis have been developed over the years. One of them, the linguistic summary, aims at summing up large volume of data into simple sentences. In this paper, we present an overview of two main streams of research, namely fuzzy logic based systems and natural language generation, covering the methods designed to work with numerical data, time series, or simple labels (enumerations). We focus… Voir plus

Along with the increase of the amount of data stored and to be analyzed, different techniques of data analysis have been developed over the years. One of them, the linguistic summary, aims at summing up large volume of data into simple sentences. In this paper, we present an overview of two main streams of research, namely fuzzy logic based systems and natural language generation, covering the methods designed to work with numerical data, time series, or simple labels (enumerations). We focus on the former stream and we give some hints to go further on fuzzy quantifiers. Voir moins