Mark's Blog

I've often thought that "tagging" is just more web2.0 techno-babble hype. People have been doing it since the days of library index files (where the topic index had to be maintained by hand), then later with HTML keywords (which proved useless). Tagging solves the one problem of finding many items of similar topic from among a larger but contained group of items, all with similar format. For example, it is perfect for finding all of Mark's recipes (or Linux hacks) or the beaches I've written about in my blog. If your base of items is large enough, it can be thought to be the equivalent of finding a book on how to administer Linux or a guidebook to Kauai.

In a way, the power of tagging does have interesting results when the base grows big enough, which is the point of the taggers, I suppose. If there were some mechanism for universal tagging, searching for "Kauai AND beaches" would reveal a virtual guidebook created from the collective knowledge of web publishers. It would be bigger and better than any single published guidebook. However, given tagging errors (and not even considerig tag spam--you heard it here first) such a tag search would realistically only be marginally better than what you get today with Googling "Kauai beaches". And frankly, I think Google still has plenty of margin to increase the relevancy of its results, which would be much easier to implement than universal tagging.

I think this is why Google won the search engine wars. After alta-vista faltered, Yahoo and Google were the main choices. I used Yahoo twice and got lost in its categories and have used Google ever since.

Everybody points to Flickr (the photo sharing website) as the model for tags. That's wrong because we do not have any way of indexing a photo other than by the words its author uses to describe it. In the Flickr model, those words are tags, eg Black&White Paris Cafe Kiss Romatic Doisneau-style. I've noticed that Flickr users over-tag their photos, such as adding Hotel-de-Ville in the previous example which is completely extraneous to actually finding similar photos. In the tag paradigm, this fortunately does no harm, although it may explain why Flickr is so slow. In the end, I think Flickr would be better off having people provide a title and literal description of their photo and let the search engines deal with it. Users can always add site=flickr.com to limit their searches--but why would they?

In the end, tags are not the way to find a specific item, and that's probably what 90% of all searchers want, for example a pizza recipe or photos of Wailua beach. The tool for that is a full-text search. In theory, a set of items perfectly tagged by its authors or readers would approach the usefulness of a search engine (since we only tend to search on keywords). But why have error-prone humans do the job, when a computer can index every word much faster, cheaper, and perfectly accurately?

As for searching email with Google, I would prefer not to trust them to keep my private email private. I would however like better search tools in Mozilla, etc. Then I only have two search domains: public and private, though realistically I have several private domains that aren't connected. Yet the work of performing multiple searches is still preferable to letting Google "own" everthing (since they're the ones that will make money off of it...).

The perfect job is what I don't think we can do. Certainly some people are very good at indexing. (I'm not.) But they cannot know in advance what someone's going to want to search for a year from now.

To me it's not the accuracy, but the low cost and speed of building indexes for everything that makes an after the fact search engine better than tagging.

Tagging may be good for pictures. Has it helped you? How do you know what words to tag a picture with?