Yahoo continues down the path of more tagging and more collaborative content. Having already purchased Flickr, this morning they’re acquiring del.icio.us (terms undislosed):

From Joshua Schachter at the del.icio.us blog:

We’re proud to…

I think you also make a good point that never followed outgoing links may have value, though I am concerned that they usually may be spam, as in the example you gave.

You mentioned an interest in personalized search here and in your previous post. This paper focused on a profile-based method of personalized search, building a list of your interests, group or individualized personalization vectors for those interests, and using that to bias all of your searches. This is also the approach described by the Kaltix team and used in Google Personalized Search.

The problems with this approach are that it is expensive to compute all the personalization vectors (or vector fragments), the personalization will not adapt quickly to new data or trends, and the personalization has to be fairly coarse-grained to have any chance of being feasible.

The approach where I have focused my attention is using short-term behavior to do fine-grained search personalization. For example, if I do search A, don’t find what I want, then refine that search to search B, the two searches are treated independently. I see the same results for search B as everyone else sees. That is clearly wrong. There is valuable information in what I found or failed to find in search A that should be applied to improve the results in search B.

More generally, the search and clickstream history of each user seems like it should be part of computing the relevance of the search results for that user.