Skip to Main content Skip to Navigation
Theses

Découverte et enrichissement de connaissances à partir de textes pour la recherche d’experts

Abstract : Expert finding consists in the identification of a set of individuals who are considered tobe experts in a particular topic. This is an essential problem in the academic world. Indeed,it is constantly necessary to identify suitable researchers when setting up reading orevaluation committees for research projects, for example. Indeed, it is particularly useful toautomatically identify experts on a specific field from the scientific literature. We suggestan approach for knowledge discovery and enrichment based on a semantic annotation ofscientific articles, on their representation in the form of scientific collaboration networksand their exploration using a graph abstraction method. This method makes it possibleto focus on dense areas of networks and to discover experts and their associated expertiseusing connectivity constraints. The latter make it possible to take into account a validationby peers, materialized by the density of scientific collaboration relations that individualsmaintain with each other. We test our approach on a corpus of scientific publications,propose an original method for evaluating our results and compare our performance toexpert research methods implemented in the LT ExpertFinder evaluation framework. Weobtain better performance than the state of the art and discover that the most decisiveindicators of expertise are the writing of highly cited articles but also the ability to citethe appropriate scientific literature
Document type :
Theses
Complete list of metadata

https://tel.archives-ouvertes.fr/tel-03425132
Contributor : Abes Star :  Contact
Submitted on : Wednesday, November 10, 2021 - 5:05:54 PM
Last modification on : Thursday, November 11, 2021 - 4:00:00 AM

File

edgalilee_th_2021_zevio.pdf
Version validated by the jury (STAR)

Identifiers

  • HAL Id : tel-03425132, version 1

Citation

Stella Zevio. Découverte et enrichissement de connaissances à partir de textes pour la recherche d’experts. Recherche d'information [cs.IR]. Université Paris-Nord - Paris XIII, 2021. Français. ⟨NNT : 2021PA131019⟩. ⟨tel-03425132⟩

Share

Metrics

Les métriques sont temporairement indisponibles