Preliminary Evaluation of Passage Retrieval in Biomedical Multilingual Question Answering

Question answering systems can support biologists and physicians when searching for answers in the scientific literature or the Web. Further, multilingual question answering systems provide more possibilities and flexibility to users, by allowing them to write questions and get answers in their native language, and the exploration of resources in other languages by means of machine translation. We present a prototype of a multilingual question answering system for the biomedical domain for English, German and Spanish. Preliminary experiments have been carried out for passage retrieval based on the multilingual parallel datasets of Medline titles released in the CLEF-ER challenge. Two parallel collections of 50 questions for English/German and English/Spanish have been created to support evaluation of the system. Results show that the task is not straightforward, that additional semantic resources are needed to support query expansion and that different strategies might be necessary for distinct languages.