Tags are no good at all for conventional categorization – mainly because the consensus on usage is not well defined.

However, I believe that tagging can take on two purposes (which each have their own pros and cons). Many people who use tags fall into one or the other.

First, that of a personal bookmark system. In this way, tagging need only make sense to the user. This is an extreme advantage because one can find bookmarks or classification from their own choosing.

Secondly, tagging can be seen as a classification system for general, or group, use. Tagging must be conventional in this domain for it to make sense for the group as whole.

I wrote more of my thoughts on this issue here.

This week we saw a number of interesting web 2.0 developments. Far outstripping everything else, however, was the terrorist attack on London on July 7. Blogs and the Internet played a big role in getting information out to people, as television lagg…

Publisher’s have an incentive to tag and tag correctly, but there is also a large incentive to spam. Automated systems can help with this, but spammers will always get through when the data is so thin.

User tagging is (or can be) much deeper for any given URI. It seems like this is what you are focusing on above. There are many ongoing experiments here and we’ve written about a bunch of them at TechCrunch. The bottom line is that the users need a good incentive to tag correctly, and you can get very good data by combining the tagsets. Services that have a good user incentive include delicious and furl (people want to find this stuff again later). Feedster’s current solution seems to lack any user incentive whatsoever, and so very bad (and/or very little) tag data is created. So good user incentive = possibly good data. No user incentive = bad or no data.

I also think that there are simply too many new tagging “communities” sprouting up and that the masses won’t want to use many of them. They must find a way to interconnect or most will fail to generate deep enough data.

I bet delicious is generating some great tag data though that could feed one heck of a search engine one day…

If you are trying to help the average user, then the “bad” examples that you listed do not apply.

- Experts can never be better at saving my information. This is something only I can do. If I save a link on blinklist for example, then only I know what tags I want to use. We can use an auto-suggestion script to make the task easier for the user but experts don’t know what a certain site means to me.

- While tags can be too focused, that is their inherent advantage. When I blinklist a site, then the tags are my mental notes and pictures of what I remember.

True, when I try to share links with many users, we all have to agree on the same tags so it can get more difficult for a broader group. However, even here search filters can help point users in the right way and tagging is far more flexible and scaleable. After all, dmoz never scaled.

- Yes, dynamic information might change and make tags obsolete. This is true but the same would apply in an expert system. Someone would have to update this information.

- As long as I use tags for myself or withing a set group of users (either a community or a social network) then they cannot get spammed. The spamming will happen if you open it up broadly and sure enough, there are people spamming blinklist. This is a challenge and will hurt the utility and usefullness of blinklist as a broader discover engine.

This is perhaps one of the toughest challenges to solve. While tagging is great for personal and group sharing, retrieving, and finding (knowledge management) the utility for broader discover will only work if tagging gains critical mass and the discover engines figure out how to keep listings clean and the spammers out.

Wikipedia has managed to maintain a user generated site pretty well. Perhaps the tagging sites will find a way to solve the problem too.

As such, I think that it has good points and bad. But what if that isn’t the main purpose of tagging? What if the main purpose of tagging is allowing a dispersed group of loosely coupled users to bring together a wide-variety of information and consolidate, via the tags, into a single attention stream? That is a totally different use than classification etc and would result in a very different set of benefits and drawbacks.