Latent Semantic Indexing

Latent Semantic Indexing [LSI] has no impact on SEO - and here's why..

When a whole bunch of patents were filed by Google in 2006, all around information retrieval and associated capabiities, eyes were on guard. Whenever Google does anything - SEO Chasers are in hot pursuit, to figure out the impact on their website rankings.

To prove LSI is NOT having any effect on your Search Rankings. Lets look at the evidence - that which we can measure.

Google does not own or use any patents on LSI - as far as can be established. It uses its own proprietary algorithms for search and Adwords functions.

LSI relies on the way different words, used in the same context, to communicate the same meaning are used, and how often. For instance, car and automobile can be used interchangeably in most instances, as can cellphone and mobile phone. In the past, search engines could not make this distinction and pulled results for each term independently. Today, LSI provides a framwork upon which such word groups can be associated and returned in search results to give broader output. Over the 50 billion pages on the Net, compiling these relationship matrices is quite a task.

So frequency will start to dominate. The words are decomposed into around 300 rows using Singular Value Decomposition. This significantly smaller subset of words represent all thousands of rows previously identified.

Words are then scored based on 'similarity' in meaning to the words in the 300 rows.

LSI works with words at the level of 'meaning', rather than the group of letters that make up words. It also deals with different tenses of the same word, and plurals of words. In spite of this, Google results for plurals typically differ by over 20% from the singular version, and search results on the first page are noticeably distinct. Not something that indicates the use of LSI. Instead Google has a better system, taking into account proximity of content across billions of documents.

LSI is based on a rather loose academic theory. Google search uses meaninful human related context - based on human behaviour and common relationships.

If you want a real sharp SEO Strategy beyond keyword density, anchor text, backlinks etc, take a look at Referential Integrity [RI]. RI works with Google, rather than against it.

NEXT: Referential Integrity

Back To Top

SEO Index | Search Engines | Google Page Ranking | SEO Testing | Optimization | Keywords | Using Videos | Using Sitemaps