Section: Overall Objectives

Highlights of the Year

• In Nguyen Khang Pham's thesis, we tried to use a classical data analysis tool, the factorial correspondance analysis, with images. This approach was very successful: The method appeared to be more efficient than the techniques used like LSA and pLSA or the TF.IDF weighting scheme. Moreover, these algorithms were developed for GPU chips and are very fast. See section  6.1.2

• We have launched in 2008 a new collaboration with the PILGRIM team of IRISA whose goal is to revisit the principle of search engines based on the fuzzy set theory. In this framework, queries and document are fuzzy set sets of words which are paired by the use of gradual implications. This work provided first very promising results and was published in ECIR 2009 and Fuzz-IEEE'09, two major conferences in respectively Information retrieval and Fuzzy Logic, a proof that the work, although still preliminary, is of interest for the community. See section  6.2.2