One of main problems in Cross-Language Information Retrieval (CLIR) in how to cross languages. This report describes the work done by the IRIT team in the multilingual, bilingual and monolinngual tasks at Cross-Language Evaluation Forum (CLEF). Our approach to CLIR is based on query translation. The queries were translated using free dictionaries and then two disambiguation strategies were compared. The first is based on aligned corpora and second on bid-directional translation. The experiments were performed using the marcure system