For some time now search engines like Google and Yahoo have been employing Latent Semantic Indexing (LSI) as part of their search results. Google started this back in 04.
While LSI is not a new search engine optimization topic, it is very important in developing relevancy for ranking, building out the long tail of search but now also for universal search.
What is LSI and how does it work? Latent Semantic Indexing is a key component in how the search engines understand the “context” of website content. LSI is based on a concept called co-occurrence clustering of words and the association of documents based on the occurrence data.
In short, the search engines essentially learn what words mean based on the other words around them. This practice is particularly important in many languages as it is fairly common for words to have multiple meanings.
For example: let’s look at the word “shoes”.
At this point, you really don’t know what kind of shoes I am referring to.
If the surrounding text had content that described triathlons, marathons, jogging or features brands like Nike, Brooks and Saucony, now you know that shoes means running shoes.
However, if the surrounding text stated drums, pads, rotors, disc, and metallic or features brands like Raybestos and Bendix, now you know that shoes means brake shoes.
Of course, if the surrounding text stated hoofs, paddocks, stables, and horse stalls, now you know that shoes means horseshoes.
Having these other related words on a page helps the search engines understand the “context” of a given keyword and support the relevancy of the search result. Without these clarifying words, you do not know what kind of shoes will arrive in your search query. With these related words, now the search results are more relevant.
In addition, these related words helps build the long tail of search keywords. When keywords and content are developed appropriately, it is not uncommon to end up with several hundred to several thousand keywords. In fact, many of our clients often have but 25 keywords as part of their SEO campaign but 12 months later have 500-2,500 keyword combinations and permutations gathered via Google Analytics.
Now with universal search, LSI is even more important. The purpose of universal search is to help identify and refine search intent. Universal search is now associating different media types as products of the search result.
If you search for Hawaii Vacations, people expect to pictures if not video of a beach on Maui. Not only do you want to read a review of a specific tour, but you want to hear a podcast of that tour.
If you sell Plus Size Prom Dresses, people not only expect to see last years styles but this years styles. If you are looking for a prom dress, chances are you also need matching shoes, handbag, hair accessories, etc.
In addition to pictures, search engines like Google and Yahoo know that people that are interested in prom dresses also frequent social media or networking sites like YouTube, MySpace, FaceBook, MyYearbook, Imeem, Bebo and Flickr. As a result as a marketer, if you sell prom dresses, you should have a MySpace page geared for teenagers. Perhaps you should upload several prom music playlists on Imeem or post on their blog. Better still, create a Flickr account where you encourage teenagers to share their prom photos wearing the prom dress they bought at your store. This is another application of LSI where pages of related topics being linked together can strengthen or help define the “meaning” for phrases on the pages you want to get ranked.
External blogs are another application of LSI. Blog posts may link your home page, an interior page or even your blog. External blogs increase exposure to your website and reinforce the main theme and keywords.
LSI is as important as ever in search engine optimization. If you are interested in the math behind LSI, there are several sources. The Journal of Graph Algorithm and Applications has published a series of papers on LSI.
Others at universities like Penn State and the University of Colorado at Boulder have written some interesting papers and applications on this topic.
No comments:
Post a Comment