Matching Fusion with Conceptual Indexing

Karam Abdulahhad (1) , Jean-Pierre Chevallet (2) , Catherine Berrut (2)

1 LIG - Laboratoire d'Informatique de Grenoble
2 MRIM - Modélisation et Recherche d’Information Multimédia [Grenoble]

Many studies have been addressed the term-mismatch problem, which arises when using different terms or words for expressing the same meaning. We also introduce another problem: over-specialized document, which is caused when IR systems prefer documents that have poor query-document intersection, but with high weighting value, to those that have rich query-document intersection with low weighting value. In this study, we propose to use, simultaneously, multiple types of indexing elements: ngrams, keywords, and concepts, instead of only keywords. We followed a late data-fusion technique to achieve that. Through our proposed model, we also try to overcome the over-specialized document problem. Experiments for model validation have been done by using ImageCLEF2011 test collection, UMLS2009 Meta-thesaurus, and MetaMap tool for mapping text into UMLS concepts.

Domaines: Recherche d'information [cs.IR]

Karam Abdulahhad, Jean-Pierre Chevallet, Catherine Berrut. Matching Fusion with Conceptual Indexing. RISE 2012 - Atelier Recherche d'Information SEmantique (associé à la conférence EGC 2012), Jan 2012, Bordeaux, France. pp.34-45. ⟨hal-00770561⟩