Kaltix and personalized search

There are some interesting rumors floating around about Kaltix, a stealth start-up out of the Stanford WebBase Project. This is the same group that created the PageRank algorithm that was later spun out as a little start-up called Google. As you might expect with a company in stealth mode we’re still long on speculation and short on facts, but it looks like their main technology is a faster way to compute PageRank, the algorithm used by Google to rank hits from a search based on the Web’s link structure.

This is interesting because it would allow Google (or any other search engine) to quickly recalculate personalized indexes for each and every user. After seeding a personal index with my bookmarks file, Google would know that when I for “Jaguar” I’m probably interested in the latest version of Apple’s OS, not the car or the cat. The CNET article has a good overview, but Jeffrey Heer’s blog has a nice perspective as a researcher who happens to be housemates with one of the Kaltix founders.

There are a lot of question-marks still, and I’m not yet convinced that Kaltix’s technology is the crown jewels that Heer or the CNET article claim it is. Speedy indexing is necessary for large-scale personalized search, but you still need to create a profile from something. The real question will be whether a search engine can generate a personal profile that helps disambiguate the searches people make in actual use. Add to this the need to keep personal information like browser history from being transmitted to outside companies and you have a tall order. I’m not saying these problems can’t be solved, but as far as I know they haven’t been solved yet. I expect Kaltix will get bought by one of the big search companies, but it will still be several years before we see personalized search running on any large (non-intranet) scale.