Analysis of retrieval models for cross language information retrieval

Abstract

There are several information retrieval systems widely in use today. There are systems based on probabilistic models of relevance, language modeling and those based on DFR model. Though most of these can be applied in a cross-lingual environment, the retrieval efficiency varies widely in such a setting. In this paper we evaluate the performance of the IR models using different empirical parameters using English-Hindi corpus. We analyze the dependency of different models on the value of the empirical parameter and present the results. We observe that language modeling gives comparably better results even when relevance feedback is not available.