Enhancing search results with metadata. My tag fever has now infected the InfoWorld site, as Chad Dickerson and Matt McAlister have recently discussed. This opens up a bunch of possibilities. Today, for example, I tweaked my experimental InfoWorld power search
to report the InfoWorld-assigned tags for URLs in the result set. Only
the most recent items carry tags. But when they do exist, it seems
really useful to surface them in a search context. Since new stuff is
being tagged, and since search is weighted for recency, they'll be
showing up a lot more as we go forward.
Enhancing search results with metadata is something I've been thinking about, and doing, for a long time. I devoted a chapter of my 1999 book to the topic, and outlined the same strategies in a column way back in 2000.
The basic idea is really simple. When you're scanning a list of search
results, it's expensive to click through to a dead end. To avoid that
outcome, you'd like the results page to decorate each item with clues
that help you decide whether it's worth investing some of your scarce
attention. I think the tags will turn out to be a great source of these
clues. So will aggregated metadata from del.icio.us — common tags,
number of citations — but that'll require more caching of del.icio.us
data than I'm doing now.
You may also notice that InfoWorld's result URLs are starting to have a new look. What would once have been:
IP telephony is still about the money
would now be:
IP telephony is still about the money | InfoWorld | News | 2001-05-28 | by Stephen Lee
Surfacing these extra bits of metadata — the publication, the story
type, the date, the author — is a great way to make search results
more scannable. The mechanism is the humble, but often overlooked, HTML
doctitle — that is, the contents of your HTML pages' <title>
tags.
If you're not expressing your doctitles as a virtual metadata
repository, you're missing out on some low-hanging fruit. The strategy
pays multiple dividends because doctitles show up everywhere: in your
own search engine, in Google and friends, in local and shared
bookmarks, maybe even (as in our case) in your web analytics. [Jon's Radio]