Longtime search engine wonks will remember Gigablast from back in the day as one of the lesser-known search engines with a huge, huge page index and a scrappy founder named Matt Wells. If you weren’t looking at search engines too often then, here’s some backstory.
Gigablast is still around, but Matt is focusing on a new project now with the recent launch of a site called FlurbiT, available at http://flurbit.com/. FlurbiT bills itself as “the largest event search engine in the U.S.”, which it probably is, considering that FlurbiT mines the open, unstructured Web for event information.
From the front page you can go and browse for available events, or you can do a search for specific types of events within a given radius of a place. I decided to look for daily events within 30 miles of Oklahoma City, Oklahoma. Once you run a search you like, you can save it as an RSS feed.
I did not get a result count, but the results were broken out by date. Because I searched for things that were happening on a daily basis, there were a lot of repeats. Samples of mined events included a day camp, several zoo presentations, and an exercise program.
Almost immediately the issue of the mined data became apparent. For example, a daily exercise event does occur — but is open only to members of Absentee Shawnee Tribe, tribal employees, and Native Americans in surrounding counties. A daily appearance by Ben “Cooter” Jones actually took place in 2007. It would be good if there was an easy way to report when events have expired, are restricted in certain ways, or are otherwise different from their single-line presentation.
I tried another search, this time for events occuring weekly within 30 miles of Decatur, Illinois. This group of results was much better with events including Toastmasters, Church events, 8-ball tournaments, and martial arts classes. There were still erroneous/odd listings, however.
Each listing has a page showing the relevant, mined data with a map, and a highlighted version of the original event page so you can easily see the context for the event.
Should you find a set of event search results you really like, you can get a widget to add them to your own Web site. And you don’t have to rely on FlurbiT’s mining to get your event right; you can submit an event instead.
Trying to datamine the unstructured Web is a difficult, thankless job, and it shouldn’t be a surprise to anybody that there are erroneous results on FlurbiT. Having a simple mechanism to report incorrect listings would help a lot, and remove the chaff from what is already a large, potentially useful database of events.