Cross-language Information Retrieval by Reduced k-means

Abstract

Cross-language information retrieval aims at retrieving relevant documents in one language for a query set in another language. Here we propose a new approach to the problem of cross-language information retrieval based on factorization of a term-document matrix by an iterative method of Reduced k-means clustering. Method of Reduced k-means intended at simultaneous reduction of objects (documents) and variables (index terms). Proposed method is compared to standard machine learning techniques of cross-language information retrieval by usage of latent semantic indexing and canonical correlation analysis. Motivation for usage of Reduced k-means method for a task of cross-language information retrieval comes from an observation that documents in a semantic space obtained by method of latent semantic indexing are clustered by their language and not by their topics in the first place. As Reduced k-means aims at preserving clustering structure of data, the idea is that the proposed method could address the mentioned problem.