LIMSI@ CLEF eHealth 2015-task 2.
Résumé
This paper presents LIMSI’s participation in the User-Centered Health Information Retrieval task (task 2) at the CLEF eHealth 2015 workshop. In our contribution we explored two different strategies to query expansion, i.e. one based on entity recognition using MetaMap and the UMLS, and a second strategy based on disease hypothesis generation using self-constructed external resources such a corpus of Wikipedia pages describing diseases and conditions, and web pages from the Medline Plus health portal. Our best-scoring run was a weighed UMLS-based run which put emphasis on incorporating signs and symptoms recognized in the topic text by MetaMap. This run achieved a P@10 score of 0.262 and nDCG@10 of 0.196, respectively.
Origine | Fichiers produits par l'(les) auteur(s) |
---|
Loading...