Does IBM's Watson Offer the Future of Search? Not Yet
In the past month, some leading voices have disparaged Google (GOOG), Microsoft's (MSFT) Bing and other search engines for producing results they claim are polluted with spam from content farms like Demand Media (DMD) and Yahoo's (YHOO) Associated Content. Some of these voices hailed the victory of IBM's Watson over human Jeopardy! champs as a forerunner of a new type of more intelligent, useful and less spammy search.
Prominent technology guru and author Vivek Wadhwa spins a vision for a search engine that would be more like a really useful Q&A service that knows your friends, your preferences and your location and can respond quickly and accurately to your queries as they relate to you, personally.
In Wadhwa's world, these new and useful search offerings would rely heavily on a searcher's "social graph." The theory is simple: Trusted recommendations and collective knowledge from groups directly associated with a person should return better information than recommendations and knowledge from random sources, such as Demand Media or Associated Content.
Oodle.com founder Craig Donato has commented on the rise of social commerce, a world where purchasing decisions about goods and services are most strongly influenced by recommendations from those nearest to you in the social graph. This echoes what Web 2.0 Conference curator and search expert John Battelle has long called the Semantic Web. I've also covered peer-to-peer search engine Wowd, which in theory could go a long way toward Wadhwa's vision.
Quickly Diminishing Returns
I have to admit, though, while I find the concept intriguing and possibly even Utopian, I don't think it will happen. My reasoning? TripAdvisor and Yelp. Both are extremely popular Web properties for reviews, probably the two leaders. Both have done excellent jobs of aggregating user-generated content. But for me, each has a key problem that points to the potential difficulties of leveraging the social graph for search.
Sponsored Links
TripAdvisor is by far the largest repository of hotel and travel reviews. And it works well for really large hotels and really popular destinations. But once you look at reviews of smaller hotels and lesser-known destinations, the number of current reviews often falls off precipitously. And what are the chances of someone in your social graph having made those reviews or even having visited those places? Probably not very high, considering that TripAdvisor has such low response rates to off-the-beaten-path locations.
Likewise, Yelp reviews don't match up with my social graph. And even more important with restaurants, what if that person who is in your social graph doesn't share your tastes? This limits the relevance of the social graph-related reviews even more.
The upshot? In my mind, the perfect search will continue to be an amalgam of curated content, professionally produced content and, to a greater degree, social content coming from a searcher's social graph. Results from the content farms will start to get screened out more aggressively as the incumbent search engines see this as essential to improving search quality.
So, Watson is an interesting exercise, and it may offer in a raw form the logic and technology side of a vastly improved search engine. But we've got a ways to go before the social graph can be the primary power behind search.
Get info on stocks mentioned in this article:
