Latent Semantic Indexing

Latent semantic indexing - indexing search results based on the pairing of like words within a document. A returned result may not even have the search term in it. It may be returned based soley on the fact that it contains many similar words to those appearing in search results which contain the search word.

Semanic organization aims to understand information based on the natural patters in language. Generally the process consists of removing stop words, stemming, and then using either singular value decomposition or muti dimentianal scaling. In large scale search engines MDS is more practicle since it is only one of various feedbacks and needs only to be almost correct. SVD is much more exact and requires much more processor power.