Gilles Hubert
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Gilles Hubert has a total of 59 co-authors. Here are the top co-authors by number of shared publications:
- Guillaume Cabanac
IRIT, University of Toulouse, Toulouse, France
13 shared publications - Christian Sallaberry
LIUPPA, E2S, University of Pau and Pays Adour (UPPA)
6 shared publications - Karen Pinel-Sauvagnat
Université Paul Sabatier, IRIT
5 shared publications A. Le Parc
3 shared publicationsClaudio Gutiérrez-Soto
Université de Toulouse, France & Universidad del Bío, Chile
3 shared publicationsDamien Palacio
University of Zurich, Switzerland
3 shared publications
Latest Publications
- Open Access
- 13 October 2026
- Published by Association for Computing Machinery (ACM)
- p. 1-11
EditorCheck: Auditing editor–author conflicts of interests
- 13 October 2026
- Published by Association for Computing Machinery (ACM)
- p. 1-11
Open AccessEditorCheck: Auditing editor–author conflicts of interests
- 15 September 2025
- Published by Springer Nature
- p. 433-451
EUR-Lex-Triples: A Legal Relation Extraction Dataset from European Legislation
- 15 September 2025
- Published by Springer Nature
- p. 433-451
EUR-Lex-Triples: A Legal Relation Extraction Dataset from European Legislation
- Open Access
- 10 April 2023
- Published by Association for Computing Machinery (ACM) in ACM Transactions on Information Systems
- Vol. 41 (4) , 1-33
TSSuBERT: How to Sum Up Multiple Years of Reading in a Few Tweets
The development of deep neural networks and the emergence of pre-trained language models such as BERT allow to increase performance on many NLP tasks. However, these models do not meet the same popularity for tweet stream summarization, which is probably because their computation limitation requires to drastically truncate the textual input. Our contribution in this article is threefold. First, we propose a neural model to automatically and incrementally summarize huge tweet streams. This extractive model combines in an original way pre-trained language models and vocabulary frequency based representations to predict tweet salience. An additional advantage of the model is that it automatically adapts the size of the output summary according to the input tweet stream. Second, we detail an original methodology to construct tweet stream summarization datasets requiring little human effort. Third, we release the TES 2012-2016 dataset constructed using the aforementioned methodology. Baselines, oracle summaries, gold standard, and qualitative assessments are made publicly available. To evaluate our approach, we conducted extensive quantitative experiments using three different tweet collections as well as an additional qualitative evaluation. Results show that our method outperforms state-of-the-art ones. We believe that this work opens avenues of research for incremental summarization, which has not received much attention yet.
- 10 April 2023
- Published by Association for Computing Machinery (ACM) in ACM Transactions on Information Systems
- Vol. 41 (4) , 1-33
Open AccessTSSuBERT: How to Sum Up Multiple Years of Reading in a Few Tweets
The development of deep neural networks and the emergence of pre-trained language models such as BERT allow to increase performance on many NLP tasks. However, these models do not meet the same popularity for tweet stream summarization, which is probably because their computation limitation requires to drastically truncate the textual input. Our contribution in this article is threefold. First, we propose a neural model to automatically and incrementally summarize huge tweet streams. This extractive model combines in an original way pre-trained language models and vocabulary frequency based representations to predict tweet salience. An additional advantage of the model is that it automatically adapts the size of the output summary according to the input tweet stream. Second, we detail an original methodology to construct tweet stream summarization datasets requiring little human effort. Third, we release the TES 2012-2016 dataset constructed using the aforementioned methodology. Baselines, oracle summaries, gold standard, and qualitative assessments are made publicly available. To evaluate our approach, we conducted extensive quantitative experiments using three different tweet collections as well as an additional qualitative evaluation. Results show that our method outperforms state-of-the-art ones. We believe that this work opens avenues of research for incremental summarization, which has not received much attention yet.
- 1 November 2019
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1-8
Comparing the Effectiveness of Query-Document Clusterings Using the QDSM and Cosine Similarity
Typically, approaches based on query clustering in IR only take the keywords that belong to the queries into consideration, with the aim of calculating the similarity among them. In search engines, one of the factors that affects precision is the short lengths of queries. One way to tackle this problem is through the use of the context. Aiming to consider the context of queries, several measures have been proposed to calculate their inter-document similarity, which are known as Query-Sensitive Similarity Measure (QSSM). This paper presents a new QSSM called Query-Document Similarity Measure (QDSM), which is used to build clusters of similar search results. The goal of this measure is to provide clusters of better quality (i.e., clusters with more relevant documents) taking into account the context in which the queries are submitted. To achieve this goal, several experiments were carried out using four well-known algorithms in Information Retrieval (IR) along with the AOL Query Logs Dataset (AOL). Final results using Bisection K-Means and Average link show that the clusters built with QDSM provide better effectiveness than the clusters built with Cosine Similarity (Sc).
- 1 November 2019
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1-8
Comparing the Effectiveness of Query-Document Clusterings Using the QDSM and Cosine Similarity
Typically, approaches based on query clustering in IR only take the keywords that belong to the queries into consideration, with the aim of calculating the similarity among them. In search engines, one of the factors that affects precision is the short lengths of queries. One way to tackle this problem is through the use of the context. Aiming to consider the context of queries, several measures have been proposed to calculate their inter-document similarity, which are known as Query-Sensitive Similarity Measure (QSSM). This paper presents a new QSSM called Query-Document Similarity Measure (QDSM), which is used to build clusters of similar search results. The goal of this measure is to provide clusters of better quality (i.e., clusters with more relevant documents) taking into account the context in which the queries are submitted. To achieve this goal, several experiments were carried out using four well-known algorithms in Information Retrieval (IR) along with the AOL Query Logs Dataset (AOL). Final results using Bisection K-Means and Average link show that the clusters built with QDSM provide better effectiveness than the clusters built with Cosine Similarity (Sc).
- Open Access
- 1 March 2018
- Published by Elsevier in Information Processing & Management
- Vol. 54 (2) , 252-272
TournaRank: When retrieval becomes document competition
- 1 March 2018
- Published by Elsevier in Information Processing & Management
- Vol. 54 (2) , 252-272
Open AccessTournaRank: When retrieval becomes document competition
- Open Access
Everything You Always Wanted to Know About TREC RTS* (*But Were Afraid to Ask)
The TREC Real-Time Summarization (RTS) track provides a framework for evaluating systems monitoring the Twitter stream and pushing tweets to users according to given profiles. It includes metrics, files, settings and hypothesis provided by the organizers. In this work, we perform a thorough analysis of each component of the framework used in 2016 and 2017 and found some limitations for the Scenario A of this track. Our main findings point out the weakness of the metrics and give clear recommendations to fairly reuse the collection.
Open AccessEverything You Always Wanted to Know About TREC RTS* (*But Were Afraid to Ask)
The TREC Real-Time Summarization (RTS) track provides a framework for evaluating systems monitoring the Twitter stream and pushing tweets to users according to given profiles. It includes metrics, files, settings and hypothesis provided by the organizers. In this work, we perform a thorough analysis of each component of the framework used in 2016 and 2017 and found some limitations for the Scenario A of this track. Our main findings point out the weakness of the metrics and give clear recommendations to fairly reuse the collection.
- Open Access
- 1 May 2017
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 221-232
Expert suggestion for conference program committees
With the rapid development of science, the academic community requires higher and higher quality of the published articles. This great responsibility is placed on editorial boards of journals, on program committees of conferences and their members. In addition, with a large number of scientific conferences held each year, searching for experts that would be invited to join the program committees is an increasingly hard task. In this paper, we propose an expert modelling method on different indicators of expertise, in particular on citation networks in order to build suggestions of members for program committees of conferences. Contrary to most of the works about expert search in the literature which are based on textual sources of evidence to define the expertise of an expert candidate, our model leverages non-textual indicators only. We evaluate the proximity of an expert candidate according to the weights of his/her various relations with the conference for which the candidate is proposed to be member of the next program committee. We report various experiments in the context of the SIGIR conference editions. These experiments give hints on the composition of program committees and on the effectiveness of our approach.
- 1 May 2017
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 221-232
Open AccessExpert suggestion for conference program committees
With the rapid development of science, the academic community requires higher and higher quality of the published articles. This great responsibility is placed on editorial boards of journals, on program committees of conferences and their members. In addition, with a large number of scientific conferences held each year, searching for experts that would be invited to join the program committees is an increasingly hard task. In this paper, we propose an expert modelling method on different indicators of expertise, in particular on citation networks in order to build suggestions of members for program committees of conferences. Contrary to most of the works about expert search in the literature which are based on textual sources of evidence to define the expertise of an expert candidate, our model leverages non-textual indicators only. We evaluate the proximity of an expert candidate according to the weights of his/her various relations with the conference for which the candidate is proposed to be member of the next program committee. We report various experiments in the context of the SIGIR conference editions. These experiments give hints on the composition of program committees and on the effectiveness of our approach.
- Open Access
- 10 July 2016
- Published by Association for Computing Machinery (ACM)
- p. 159-168
Social Media-Based Collaborative Information Access
The notion of implicit (or explicit) collaborative information access refers to systems and practices allowing a group of users to unintentionally (respectively intentionally) seek, share and retrieve information to achieve similar (respectively shared) information-related goals. Despite an increasing adoption in social environments, collaboration behavior in information seeking and retrieval is mainly limited to small-sized groups, generally restricted to working spaces. Much remains to be learned about collaborative information seeking within open web social spaces. This paper is an attempt to better understand either implicit or explicit collaboration by studying Twitter, one of the most popular and widely used social networks. We study in particular the complex intertwinement of human interactions induced by both collaboration and social networking. We empirically explore explicit collaborative interactions based on focused conversation streams during two crisis. We identify structural patterns of temporally representative conversation subgraphs and represent their topics using Latent Dirichlet Allocation (LDA) modeling. Our main findings suggest that: 1) the critical mass of collaboration is generally limited to small-sized flat networks, with or without an influential user, 2) users are active as members of weakly overlapping groups and engage in numerous collaborative search and sharing tasks dealing with different topics, and 3) collaborative group ties evolve within the time-span of conversations.
- 10 July 2016
- Published by Association for Computing Machinery (ACM)
- p. 159-168
Open AccessSocial Media-Based Collaborative Information Access
The notion of implicit (or explicit) collaborative information access refers to systems and practices allowing a group of users to unintentionally (respectively intentionally) seek, share and retrieve information to achieve similar (respectively shared) information-related goals. Despite an increasing adoption in social environments, collaboration behavior in information seeking and retrieval is mainly limited to small-sized groups, generally restricted to working spaces. Much remains to be learned about collaborative information seeking within open web social spaces. This paper is an attempt to better understand either implicit or explicit collaboration by studying Twitter, one of the most popular and widely used social networks. We study in particular the complex intertwinement of human interactions induced by both collaboration and social networking. We empirically explore explicit collaborative interactions based on focused conversation streams during two crisis. We identify structural patterns of temporally representative conversation subgraphs and represent their topics using Latent Dirichlet Allocation (LDA) modeling. Our main findings suggest that: 1) the critical mass of collaboration is generally limited to small-sized flat networks, with or without an influential user, 2) users are active as members of weakly overlapping groups and engage in numerous collaborative search and sharing tasks dealing with different topics, and 3) collaborative group ties evolve within the time-span of conversations.
- 22 June 2016
- Published by Walter de Gruyter GmbH
- p. 101-131
Conditions et possibilités d’un « urbanisme résilient » en zone inondable
Conditions et possibilités d’un « urbanisme résilient » en zone inondable was published in Les villes à la croisée des stratégies globales et locales des enjeux climatiques on page 101.
- 22 June 2016
- Published by Walter de Gruyter GmbH
- p. 101-131
Conditions et possibilités d’un « urbanisme résilient » en zone inondable
Conditions et possibilités d’un « urbanisme résilient » en zone inondable was published in Les villes à la croisée des stratégies globales et locales des enjeux climatiques on page 101.
- 1 January 2016
- Published by IGI Global Scientific Publishing
- p. 1117-1137
On The Reuse of Past Searches in Information Retrieval
On The Reuse of Past Searches in Information Retrieval: Study of Two Probabilistic Algorithms: 10.4018/978-1-4666-9562-7.ch057: When using information retrieval systems, information related to searches is typically stored in files, which are well known as log files. By contrast, past
- 1 January 2016
- Published by IGI Global Scientific Publishing
- p. 1117-1137
On The Reuse of Past Searches in Information Retrieval
On The Reuse of Past Searches in Information Retrieval: Study of Two Probabilistic Algorithms: 10.4018/978-1-4666-9562-7.ch057: When using information retrieval systems, information related to searches is typically stored in files, which are well known as log files. By contrast, past