Xerox TREC-6 Site Report: Cross Language Text Retrieval.

This track examines the problem of retrieving documents written in one language using queries written in another language by using a bilingual dictionary at query time to construct a target language version of the original query.

Abstract

Xerox participated in the Cross Language Information Retrieval (CLIR) track of TREC-6. This track examines the problem of retrieving documents written in one language using queries written in another language. Our approach is to use a bilingual dictionary at query time to construct a target language version of the original query. We concentrate our experiments this year on manual query construction based on a weighted boolean model and on an automatic method for the translation of multi-word units. We also introduce a new derivational stemming algorithm whose word classes are generated automatically from a monolingual lexicon. We present our results on the 22 TREC-6 CLIR topics which have been assessed and briefly discuss the problems inherent in the cross-language IR task. 1 Introduction Cross Language Information Retrieval (CLIR) addresses the problem of retrieving documents written in one language using queries written in another language. As document repositories grow in size and ...