Last Friday’s announcement that Yahoo is buying del.icio.us has probably got more than a few people thinking about the future of the service and whether they want to keep using it. In any case, as with all of the interesting and useful web serv…

I think metrics around reputation and relationship to trusted sources can keep the signal to noise ratio higher, at least for using bookmarking and tagging as input to search, since spamming patterns can be identified and blocked over time.

In the limited volume of traffic I see here, there are several prominent clusters of spam commenter topics and network signatures which can be used to filter most of the noise automatically. I’m not sure what the del.icio.us team put in place a while back after spam started floating to the “Popular” list but it’s clearly helped reduce the noise. Plus, all posts are tied to a user name, providing another tool for screening.

Applications that rely on real-time or near-time input (buzz trackers, etc) have a harder problem since they generally have to make a decision on the spot.

Bookmarking and tagging by a honest user should be a useful indicator for relevance and reveal some clues about topical intent as well. Up to this point, the average del.icio.us (or any tagging service) user is generally “honest” and somewhat well informed. A rapid expansion to millions of new users on Yahoo may dilute the average quality of the incoming bookmarks, and is likely to amplify the popularity of the popular bookmarks, since they are both “interesting” and visible. It might work out that all bookmarks with more than a few votes are worth paying attention to. This would be analogous to the Bloglines metrics around the number of feeds with more than N subscribers, hardly any have more than 50 subs.

Perhaps there are natural size limits for a useful social tagging community within a given topic domain, below which there aren’t enough eyeballs to capture interesting content and provide enough votes, and above which the items, tags, and taggers need to be clustered down to manageable size.

Still thinking about this. Also continuing to look at how to manage personal bookmarking, tag and attention data, with an eye toward search. I’m sure it’s on a lot of people’s minds this weekend.

I think it is a dangerous assumption that this can continue as tagging tries to enter the mainstream. As sites like del.icio.us become more popular, the reward from spamming them increases substantially.

For more on tagging and its potential impact on search, see my previous weblog post, “Questioning tags”:

http://glinden.blogspot.com/2005/04/questioning-tags.html

Yahoo! is clearly looking to build an extend their lifestyle brand:
1. Flickr = photos
2. del.icio.us = links
3. Upcoming = events

I bet that they’ll go for Odeo next. Sure they offer a podcasting engine now - but why not incorporate an existing tool?

I agree, it will be interesting to see how this turns out.

Google has some pretty sophisticated clustering algorithms to try and do this computationally. Is there really enough user data to make a big difference? After all isn’t spending millions and millions building a user community a riskier move than using that same amount of money to research algorithms and purchase several super computers?

Google has started tagging in GMail and in Personalized Bookmarks. Not to mention Google base… I think they are waiting to see what happens before jumping in with two feet.

Also, what happens when spammers can hire thousands of people in India and China to start spamming these networks? It happened to Usenet, what’s to prevent it from happening again?

You can read more of my reponse to the acquisition in my response to the news.