UniNE at CLEF-IP 2009

Abstract

This paper describes our participation to the Intellectual Property task during the CLEF-2009 campaign. Our main objective was to evaluate different search models and try different strategies to select and weight relevant terms from a patent to form an effective query. We found out that the probabilistic models tend to perform better than others and that combining different indexing and search strategies may further improve retrieval. The final performance is still lower than expected and further investigations are therefore needed in this domain. © 2010 Springer-Verlag Berlin Heidelberg.