The use of external text data in cross-language information retrieval based on machine translation

Abstract

This paper explores the use of an external (i.e. non-target) document collection in cross-language information retrieval (CLIR) based on machine translation (MT). In our CLIR and monolingual IR experiments using an external target language collection, we show that parallel pseudo-relevance feedback is comparable to collection enrichment. In our CLIR experiments using an external source language collection, we show that context-sensitive translation of pre-translation expansion terms outperforms word-by-word (or context-free) translation on average. Moreover, we show that the combination of context-sensitive translation with pseudo-relevance feedback significantly outperforms the corresponding context-free combination and the pseudo-relevance feedback component. Thus, context-sensitive translation for pre-translation expansion is probably superior to context-free translation.