Building Proximity Models for Cross Language Information Retrieval

Abstract

In information retrieval systems, the proximity of query terms has been employed to enable ranking models to go beyond the ”bag of words” assumption and it can promote scores of documents where the matched query terms are close to each other. In this article, we study the integration of proximity models into cross-language information retrieval systems. The new proximity models are proposed and incorporated into existing cross-language information systems by combining the proximity score and the original score to re-rank retrieved documents. The experiment results show that the proposed models can help to improve the retrieval performance by 4%-7%, in terms of Mean Average Precision.