LSA
Latent Semantic Analysis (aka Latent Semantic Indexing) (LSA/LSI) is a mathematically based machine-learning method for acquiring and representing the central meanings of words and documents and for using the derived representations to enable information retrieval and analysis. An important property of LSA is that, because it represents the similarity of meaning of any two words on which it has been trained, it can also represent the meaning of any two documents as similar or different independently of the literal words they contain. As such, it provides semantic search capabilities across a document collection based on concepts and semantic understanding instead of lexical matching. More information
The possible applications for LSA technology are numerous, but the processing required to support LSA is complex and computationally demanding. SBTC has been helping clients deploy LSA technology in domain specific applications for several years. This work has included the refinement of key LSA computational routines, research in cross-language information retrieval using LSA, and consultation supporting both educational research and commercial application of LSA technology. Dian Martin, one of the principals of the company, has authored several papers and book chapters on the subject.
SBTC has the expertise and experience to overcome the various obstacles of scalability and complexity surrounding the use of LSA and can support all the ancilary pre- and post-processing tasks necessary to apply this technology to various application domains. Our strength in Numerical Computation and Parallel Algorithm Development has enabled us to develop an optimized scalable LSA/LSI implementation. We are currently preparing a commercial release of our LSA Toolkit - details to be announced.
Contact us to find out how we can help you utilize LSA technology in your application domain.