The limitation of the SVD for latent semantic indexing

Latent semantic indexing (LSI) is an indexing method for improving retrieval performance of an information retrieval system by grouping related documents to the same clusters so that each of these documents indexes the same (or almost the same) words, and unrelated documents index (relatively) different words. The de facto standard method for LSI is the truncated singular value decomposition (SVD). In this paper, we show that the LSI capability of the truncated SVD is not as conclusive as previously reported; rather it is a conditional aspect and when the condition is not met, then the truncated SVD can fail in recognizing the related documents resulting in a poor retrieval performance.