Medical Image Retrieval: ISSR at CLEF 2010

This is the second participation of Institute of Statistical Studies and Research (ISSR) group in CLEF 2010-Medical image retrieval track. This paper describes our experiments in monolingual and multilingual tasks. First, we test Paragraph Extraction (PE) and Sentence Selection (SS) approaches on the classical medical retrieval task (Ad-hoc), as well as on Case-based retrieval. Second, we compare between three Cross Language Information Retrieval (CLIR) methods. These methods are Machine Translation (MT), dictionary translation as well as translating via thesauri. For indexing and retrieval, we used the Lemur toolkit. Regarding ad-hoc retrieval task best results obtained when image caption and title used only, and for case-based task, there is no significant difference between adding extra text to the article and using its title and its image captions. For multilingual task, there is no significant difference between the three methods.