Latent Semantic Indexing (LSI) and How Using Related Words in Content Brings Positive Ranking Results

By Howard C. Gray

"Google recently strongly integrated latent semantic indexing into their algorithm." SEO Book February 2007.

In breaking down the sentence of Latent Semantic Indexing (LSI) we get the following. Latent: Hidden. Semantic: Meaning in language. Index(ing): a list, reference or indicator of something. Therefore the meaning becomes: A hidden meaning from language that is placed in a list or reference (database in this case). Basically this long-winded description and it's acronym refer to the analysis of a document using a mathematical/statistical technique for extracting and representing the similarity of meanings of words and passages by means of a matrix system.

This matrix system is based on mathematical formulas called algorithms that separate out common words and articles such as: the, a, and, and other unimportant words in the document. From there, it continues to use different planes of the matrix (usually 2 or 3 planes or angles) to continue to refine the meaning of the page, removing unnecessary words and regrouping others that appear related.

By cross-referencing these matrix's, a mathematical formula that is software driven can scan the page and find related words and thereby "determine" what a page is about. For example, if a page was written about Nuts but the word Nut(s) does not literally appear on the page, the matrix formulas look for synonyms that are similar to, or related to, the word Nuts. Pecans, walnuts, cashews and peanuts (though not a nut, actually a legume) are all defined as belonging to the category of Nuts.

Therefore if a searcher types in "Nuts" in the search box of a search engine, the incorporated LSI will find the page, determine it's about Nuts and rank the page in it's index. Essentially, think of LSI as a super Thesaurus capable of matching keywords with desired documents and web pages.

Because people use their memory to relate a word to another word, LSI is a very human approach to the very un-human Internet experience. It's also an excellent approach to acquiring strong rankings without ticking off the search engines or making your website look idiotic and/or over-optimized. We can use this technique to go after highly competitive keywords that are highly searched-for and obtain good rankings for those keywords (among others) without tripping filters designed to weed out overly optimized pages attempting to target these highly searched terms.

In addition, we can use this concept to create powerful call-to-action copy that appeals to a user's visualization. This is where we recommend using simple tools such as the WordNet Search Related Words Tool to determine new related (key)words. Even though they aren't used by searchers will still work to rank your page.