What Is Latent Semantic Indexing

Latent semantic indexing or LSI is an advanced technique for information retrieval that uses a mathematical procedure to extract the idea or concept from a group of text.  This information retrieval technique uses the natural language processing system known as latent semantic analysis or LSA.  LSA looks at the various relationships between a number of documents and the body of text found in them and establishes a group of concepts for these documents.  With LSI, the documents that are presented in response to a particular query do not necessarily have the exact words or phrases that the searcher has keyed in.

LSI offers a remedy to two of the most annoying deficiencies of the usual Boolean search technique.  These are the possibilities that a word has more than one meaning and several words having the same meanings.  These two problems are the usual reasons for documents or web pages appearing in the search results even if they are not relevant to the topic while certain web pages and documents that should have been included are absent. 

LSI is also useful for the automated specification of the categories for each document.  For this method, it uses sample documents as the foundation for understanding the concepts embodied by each category.  The technique used is to compare the ideas that are found in the example documents for each category with those that can be extracted from the document to be classified and placing it in those categories where the concepts match. 

Another benefit offered by LSI is that it can be used for any language because it is purely dependent on mathematical formulas.  Thus, it can extract the semantic content from the documents written in any language without the need to consult any thesaurus or dictionary.  The search can also be made in a particular language while the documents to be queried can be in another language. 

LSI is also applicable for terms that are not exactly words, such as the DNA sequences of genes.  Thus, biological and medical documents can easily be searched and categorized using LSI.  For example, LSI is capable of classifying genes based on the biological information that could be extracted from the abstracts and titles of biological databases.

LSI can also easily adapt itself to any modifications in the terminology and it can still function in spite of the presence of misspelled words, unreadable characters, typographical errors, and other types of noise in documents.  Therefore, LSI is applicable for a body of text that is the result of speech-to-text conversion programs and those that have been extracted from images by optical character recognition software. Check out http://ArticlesOnTap.com for more on this