Comparing Different Approaches to Treat Translation Ambiguity in CLIR: Structured Queries vs. Target Co-occurrence Based Selection

Abstract

Two main problems in Cross-language Information Retrieval are translation selection and the treatment of out-of-vocabulary terms. In this paper, we will be focusing on the problem concerning the translation selection. Structured queries and target co-occurrence-based methods seem to be the most appropriate approaches when parallel corpora are not available. However, there is no comparative study. In this paper we compare the results obtained using each of the aforementioned methods, we specify the weaknesses of each method, and finally we propose a hybrid method to combine both. In terms of mean average precision, results for Basque-English cross-lingual retrieval show that structured queries are the best approach both with long queries and short queries.