Clustering technology for search engines has been around for a while, sometimes called "relevance" searches or engines, and one in particular is worth taking a look at. Vivisimo has been around since 2000, a spinout of the very successful CS program at Carnegie Mellon. On an unrelated note, it's always been a mystery to me that the Pittsburgh area has not benefited more from the technology chops of CM, it's really amazing to look at the alumni directory and the companies that have spun out of that program. Clusty is a search engine built on top of the Vivisimo technology, and if you use search a lot and use it like I do I am pretty sure you will make this your default search engine. The bottom line is that this search engine chunks up the search results into categories. For example, a search for "saturn" will result in categories for the cars, the planet, and the NASA probes (Cassini and Voyager). As you can imagine, this is a really logical way to organize search results; I still can't believe that Google and Yahoo! haven't done more of this because it's really frustrating to hunt-and-peck around their search results (and sponsored links, grrrr) to find what you need and what you want. This clustering technology offers interesting possibilities for optimization, and it looks like the SEO crowd is starting to realize this. Here's an interesting post offering suggestions on how to take advantage of the clustering and reference for maximum effect.
That’s right, the clusters themselves could be potential keywords you could further optimize your site for. Using some of the clusters as well as a free keyword research tool like Yahoo!s Search Marketing Search Term Suggestion Tool. via Search Engine Journal.
UPDATE: Matt makes a really good point in the comments about using specific searches as opposed to simple keyword search. This is indeed a valid point, and I think it's fair to say that the natural limitations of current Google-style search has conditioned users to indeed be creative and specific about search terms. I decided to go back and use a better example, one that represents an actual use case. I am in the market for a new digital camera and am heavily leaning to the Canon EOS 350D. Here's the same search in Google. What I like about the Clusty search is how they separate reviews from manufacturer links, have another category for European links and a section for "Rebel" results (which is another name the camera goes by. If you click on the "blogs" tab things get really interesting by segmenting blog posts based on the type of post it is, e.g. "problem" vs. "shutter speed". Matt was right to bring this up and I hope the example I provided is better than the original one.