Latent semantic indexing (LSI) is a mathematical technique that attempts to ‘understand’ and categorize any object represented as text, including words and arbitrary character strings. It is used as a component of a Boolean Search system, such as Google’s search engine, for the indexing and retrieval of documents.
Semantics has to do with context. We typically are able to understand something because of the context in which we read it, see it or hear it. It is not an exact science but rather it is based on what seems to make sense. It helps us to understand things through associations.
LSI uses associations to group things into categories for improved indexing and retrieval. It attempts to overcome the linguistic challenges presented by synonyms (words with shared or similar meanings) and polysemys (words with more than one meaning) by establishing associations between different words used in the same or similar contexts. It is a fairly reliable but by no means fool-proof approach.
Keyword tools offered by the search engine marketing programs, for example, develop lists of suggested keyword options based on contextual associations. While they are extremely useful they have to be used as part of comprehensive keyword research and development in order to avoid using terms that don’t have the right association for what it is you offer. See ‘Keyword Tool’ for a further explanation.
