Approche statistique pour le filtrage terminologique des occurrences de candidats termes en texte intégral

Content Provider	Semantic Scholar
Author	Collados, José Camacho Billami, Mokhtar Boumedyen Jacquey, Evelyne Kister, Laurence
Copyright Year	2014
Abstract	Following (L'Homme, 2004), this paper focuses on terms variations in full text in French and more precisely it highlights the semantic ambiguity of terms occurrences with regards to a very high leveled distinction between terminological and general uses. This issue is very present especially in Humanities. For instance, we are interested in distinguishing between the terminological meaning of the term "sujet (subject)" in the phrase "le sujet de la phrase (the subject of the sentence)" (Linguistics) or "les reponses du sujet (subject's answers)" (Psychology), and the general meaning of the noun "sujet (topic)" that we may find in a phrase like "le sujet de cet article (the topic of this article)". In order to solve this problem, we assume that textual contexts around term occurrences give us relevant information on the kind of use we face, terminological or general. Our research is based on a statistical approach of the textual contexts. The proposed metrics are based on the hypergeometric distribution and the lexical specificity calculus as described in (Lafon, 1980). By using a manually annotated corpus as the training set, we build lexical profiles for each high leveled meaning of the term candidates. We use two methods which were compared to a baseline metric based on term frequency. The results we obtained are analyzed from both a quantitative and a qualitative point of view.
File Format	PDF HTM / HTML
Alternate Webpage(s)	http://wwwusers.di.uniroma1.it/~collados/papers/JADT_Filtrage%20Terminologique.pdf
Language	English
Access Restriction	Open
Content Type	Text
Resource Type	Article