Journées internationales d'Analyse statistique des Données Textuelles
7-10 juin 2016 Nice (France)
CoFiH: A heuristic for concept discovery in computer-assisted conceptual analysis
Louis Chartrand * , Jean-Guy Meunier  1@  , Davide Pulizzotto  1@  , José López González  1@  , Jean-François Chartier  1@  , Julian Trujillo Amaya  2@  , Tan Ngoc Le  1, *@  
1 : Laboratoire d'analyse cognitive de l'information  (LANCI)  -  Site web
2 : Université de Valle
* : Auteur correspondant

While conceptual analysis can be facilitated by computer assistance, the absence of proper models for concepts in text has curtailed the development of such tools. The most common heuristic, which consists in identifying keywords as canonical expression of a concept, poses problems of ambiguity and fails to retrieve most of the relevant textual data. In this paper, we present CoFiH, an algorithm that exploits topics in order to retrieve segments relevant to a given concept. It is then applied to C.S. Peirce's Collected Papers to facilitate the analysis of Peirce's concept of LAW. Compared to the baseline, CoFiH produces better recall and enables a meaningful analysis along several topics.

Personnes connectées : 1