A "stereo" document representation for textual information retrieval

Encouraged by a significant improvement over LSI (latent semantic indexing) approach in textual information retrieval of the DLSI (differential latent semantic indexing) approach which technically makes use of two term vectors for each document, we have proposed a concept of stereo, or multiperspective, document representation, which is expected to be effective for most of textual information retrieval approaches based on vector space model. We show that the new representation based on two or more “pictures” of each document taken from different view angles contributes to the enhanced performance of textual document retrieval by enhanced capability of extracting and capturing more individualistic features of the document. A Student t-test on experimental results on the standard Time and ADI corpora proves that the improvements of the retrieval performances of LSI/standard term vector algorithms based on multi-perspective document representation over those based on traditional single document representation are significant. ∗Corresponding Author