Me.dium has introduced a social search engine that blends results from a “real time” index of what people are currently viewing online with Yahoo’s full-scale web index. The result is something I’ve not seen elsewhere—and despite my long time skepticism regarding social search, this approach actually holds promise for delivering a unique view of relevant content on the web.
The current alpha version of Me.dium Social Search is obviously a work in process, with lots of rough edges and spurious results for many queries. But it also provides a glimpse of the web that you won’t find in the search results of the major search engines—at least not until they get around to indexing the content allowing it to rank (if ever).
Me.dium Social Search uses data gathered from its social browsing toolbar that lets you observe what others are viewing on the web in real time. With the toolbar, you’re limited to seeing a subset of what toolbar users are viewing. For the search engine, however, Me.dium is indexing the full 10-20 million URLs its two million users are looking at every day. This means that it’s the actual behavior of web users, rather than linkage patterns, “votes,” or other criteria that drives the relevance engine of the service.
This approach goes a long way toward solving “tyranny of the mob” effects that plague other services such as Digg, which can be gamed by users “voting” for content that they may not even have looked at. Instead, Me.dium observes the behavior of users to determine what’s popular, what’s new, and what’s relevant at this particular moment. “We believe it’s the first real-time index that’s ever been built,” said Kimbal Musk, CEO of Me.dium.
Me.dium Social Search finds content where your query terms appear and ranks the content according to a number of different variables, including things like “crowd rank” (essentially, popularity), “velocity” (increasing or decreasing traffic), when the site was last visited, and “average visit duration.” Results can literally change from moment to moment for the same query, based on how these variables are changing for each page in the index.
Me.dium Social Search will never knock Google off its throne, but that’s not the point. As Me.dium’s Musk noted in an interview, Google and the other major search engines are really good at three types of search: Navigational queries, database-driven queries (you know it’s out there, you just need to find it), and pure research (give me everything about a topic). Me.dium Social Search adds a new dimension: discovery. It’s similar to StumbleUpon, but with a much larger, fresher, and more dynamic set of data.
Me.dium is partnering with Yahoo, using its new BOSS (Build Your Own Search Service) open web services platform. Musk said that in this initial alpha version, results are mostly from Me.dium’s own index, but over time results will be blended with Yahoo’s full web index to provide much more comprehensive results. You can exclude Yahoo results by clicking the “I feel social” button rather than “search” when you enter your query. Vanessa has more about Yahoo’s BOSS in Yahoo! Lets You “Build Your Own Search Service”.
I’ve long argued that “pure” social search services are doomed to failure, due to problems with tagging, scope and scale issues, vulnerabilities to spamming and gaming, and a host of other reasons. I’ve argued that what will likely succeed is a hybrid approach, leveraging the wisdom of the crowds with the power of algorithmic search.
Me.dium Social Search is taking this approach. As with most alpha releases, results are inconsistent and can vary from really interesting to genuinely horrible, depending on what you’re searching for. Over time, as it tunes its algorithms and gathers even more behavioral data, Me.dium’s search results will almost certainly significantly improve.