Document Expansion, Query Translation and Language Modeling for Ad-Hoc IR

For the multilingual ad-hoc document retrieval track (TEL) at CLEF, Trinity College Dublin and Dublin City University participated in collaboration. Our retrieval experiments focused on i) document expansion using an entry vocabulary module, ii) query translation with Google translate and a statistical MT system, and iii) a comparison of the retrieval models BM25 and language modeling (LM). The major results are that document expansion did not increase MAP; topic translation using the statistical MT system resulted in about 70% of the mean average precision (MAP) achieved compared to Google translate, and LM performs equally or slightly better than BM25. The bilingual retrieval French and German to English experiments obtained 89% and 90% of the best MAP for monolingual English.