Cross-language information retrieval using latent semantic indexing and self-organizing maps

The present paper describes a method of fully automated cross-language information retrieval, which does not require any query translation. Namely, monolingual queries retrieve documents from a multilingual collection, which includes items from the query's source language. This is achieved by a method, which combines the construction of a multilingual semantic space using latent semantic indexing (LSI) and the clustering ability of self-organizing maps (SOM) for the generation of multilingual semantic categories. Tests on an English-Greek corpus reveal the effectiveness and robustness of the proposed method.