Constellation, le dépôt institutionnel de l'Université du Québec à Chicoutimi

Résumé

Rare diseases constitute a poorly known subject among the population. Nevertheless, despite their name, a large number of persons are afflicted by one or many of them. Research on the nearly seven thousand rare diseases is insufficient, and even if some works have been done to exploit scientific publications and extract relevant information, knowledge is very difficult to obtain for the general population. This paper presents a new system that try to address the dissemination of knowledge on rare diseases. Particularly, we focus on the task of extracting automatically symptoms of rare diseases from publications with a new approach using a Named Entity Recognition (NER) algorithm based on the numerical statistic Term Frequency - Inverse Document Frequency (TF-IDF).