Hacker Newsnew | comments | show | ask | jobs | submit login

Everyone loves to say "Delicious tags are great data for finding relevant information," but I'm not so sure -- at least for doing keyword queries, compared to web search engines. The right comparison is: is Delicious tagging really better than anchor text from the general Web? "Anchor text" means, the text in or near links pointing to the page. It's like an implicit noisy tagging system the Web provides (if you have a good crawl of it). It's a key component in web search. For example, if you Google for "python threads tutorial" you find nice stuff too.

I did hear a second-hand story about Yahoo web search, for what it's worth. When the Delicious acquisition happened, of course they were super excited for this very reason -- that Delicious should constitute a high-quality dataset (that Google didn't have!) Then they tried all sorts of things to incorporate it into their relevance ranking algorithm, but never could get it to work. I personally think Yahoo search had good relevance algorithm people, and therefore, if you believe this story, that Delicious data is not useful for web search relevance.

So I understand this story might be hard for others to verify. But I think it's reasonable to assume that if the data really was valuable, they would have used it for web search, and someone would have bragged about it at some point -- especially given the large amount of interest from the wider developer and computer science community about how potentially useful the Delicious dataset should be.




Guidelines | FAQ | Support | API | Security | Lists | Bookmarklet | DMCA | Apply to YC | Contact

Search: