Back in 1999 I started doing a bit of work with LSI. I didn't know it was LSI at the time but that is indeed what it was. I theorized that there had to be a better way to search and what do you think my work found? LSI is a wonderful thing!

The concept is simple. Words are just words but when you can relate words across documents you start to get a better picture and it is that pattern that is the key to better search.

It's hard to give a good example but for SEO purposes it works best when pages are written without SEO in mind because LSI can detect the patterns of a normal persons writing. It really is that simple.

A side note - I really don't think that folks such as Google and the others use LSI, at least not effectively. If they did then the ads for products such as AdSense would be near 100 percent perfect. The other thing that makes me think they don't use it is because they don't handle very many languages and in my testing I got near 100 percent perfect ad to content matching with hundreds of languages without having to code for each language. I don't have to know how to speak Hindi or Tagalog or even Gullah in order to have a machine understand it!

LSI is a wonderful thing :-)

JAG

Can anybody please simplyfy what is LSI? How can i use this.
We hav a site [snip]. We are doing SEO work for the last three month. But ranking is not improving.

Can anybody suggest how too improve rank in SERP.

[edited by: pageoneresults at 1:47 pm (utc) on Jan. 2, 2007]
[edit reason] Removed URI Reference - Please Refer to TOS [/edit]

Here's the most understandable paper I've seen on LSI that has some depth to it. It's extremely well written.

Patterns in Unstructured Data
[knowledgesearch.org...]

(This is the classic NITLE Yu, Payne, et al article. For those interested, old bookmarks didn't work, but this is apparently a new url. It's on a Creative Commons license, so it may appear elsewhere on the web too).

As the paper describes, LSI looks at the occurence of words in a document to help determine its meaning. Eg, eggs, bacon, and coffee might commonly occur in a document about breakfast.

It's not clear, though, that the engines are yet using LSI in algorithmic search. I've read some information retrieval experts say it's too slow and resource intensive for that. The paper suggests that LSI might be particularly helpful, though, in refining results.

I have seen some examples of AdSense matching that make me think that Google may be using LSI in targeting ads... where the ads were clearly prompted by word associations, rather than matching the actual words that appeared on the page. Google may also be using LSI in analyzing links.

...how we can use this technology in seo?

I don't think it's a "technology" that we "use." Rather, it suggests to me, anyway, an approach to writing, where we need to look at the overall vocabulary of a document or of our links, and keep this big picture in mind when creating our content.