News

Wikiseek Uses Wikipedia As a Starting Point for Search

Hey, this is a pretty neat idea. Wikiseek searches Wikipedia, but it goes beyond that search only those sites referenced by Wikipedia. As it notes on its about page, this idea helps make the site “an authoritative source of information less subject to spam and SEO schemes.” . You can try Wikiseek at http://www.wikiseek.com/ .

Unfortunately it doesn’t work if you try to break it. I started by doing a search for buy propecia . The first page of results included what looked like several spam blogs — Blogspot.com appeared to be the biggest offender from a quick scan. There were some legit results as well, but the spam results were enough to be bothersome.

I tried a non-trigger search, namely Wii. The results I got there were a lot better, with a lot of news articles and blog entries (though the results seemed to be slanted towards blog articles.) At the top of the screen there were categories by which you could narrow your search. I clicked on the “Game Controllers” category, and got a more focused set of results. I also got a set of results at the top of the page that were directly from Wikipedia. (They Wikipedia articles are delineated by a different-colored background; for a second I thought they were sponsored results.)

Other than being from Wikipedia and associated sites, the search results look like those from other search engines, with title, excerpt, and URL. The Wikipedia articles also have a link to edit the URL. I’d love to see the Wikipedia articles also have a last-updated timestamp, since that information is available for each article.

I did another search for “Mae Jemison” (first African-American woman in space). No category results for this one, and confusingly while there were Wikipedia results at the top of the page there were also Wikipedia results mixed with the regular results. The results were the most part very good; I can see that Wikiseek might be pretty good for more general searches, in a way you might also use a searchable subject index (if there were any good ones left. But I’m not bitter.)

Wikiseek leaves me with several unanswered questions: how often are the Wikipedia-linked sites being crawled? Any special syntax or plans for special syntax? Are there plans to try this same approach with a more closed Wiki like Scholarpedia? How are those spam blogs getting in there?

Overall I found the results pretty good, but I did find there were searches I could do that would definitely break the concept. Worth a try.

Categories: News