Latent Semantic Indexing for Patent Information

Latent Semantic Indexing (LSI) promises more accurate retrieval of information by incorporating statistical information on term meaning and frequency while retrieving documents as a result of a search. LSI’s precision and accuracy has been proven many times on test corpora, but the world’s patent literature poses a significant challenge in effectively implementing an LSI search engine due the size and heterogeneity of the patent corpus. Some of the factors which must be addressed to realize the goal of a more accurate patent search engine are discussed herein