latent semantic indexing. It is the understanding of words with more than one meaning, applying the correct meaning in order to understand what text is about. I will try to explain it, but please bear with my lack of vocabulary
Read each word then think what the page might be about
As the list went on, based on the words I used, you worked out that the tiger was 'Tiger Woods' and not some stripey beastie. A lot of those words were generic words like 'club, green, grass' but when they all come together, green means a section of the golf course rather than a colour, club would be weighted towards being a sports impliment rather than something you hit a person with, and handicap means a sytem of evening out a contest rather than a disability.
In its simplest form that is LSI at work.
Now links from pages about leisure wear, and golf carts become relevant links, while links from Stevos pet mart become less relevant. This is because the search engine understands what the pages are about and can compare semantic relevance it ensures that themed sites gain the most benefit from their links.
Step it on and add Trustrank, and it will be an easy move to negate the value of off topic links from trusted pages This reduces the temptation for webmasters to hack .gov sites etc it prevents them listening to the voices that say 'come to the dark side Luke'
Semantics is also what has allowed google to deliver targeted AdSense to pages. They acquired the technology a few years ago when they bought a company called Applied semantics ( SEO's knew then that semantics would eventually come to play a big part in the Google algo.