Thomas J. Casserly's 2007 Reading Notes


Web: 2007-1 Searching videos /


28 February 2007
2007-1

World Wide Web

(2007-1, February 28, 2007)
"Millions of Videos, and Now a Way to Search Inside Them"
Jason Pontin, The New York Times, February 25, 2007.
http://www.nytimes.com/2007/02/25/business/yourmoney/25slip.html?_r=1&oref=slogin

Keywords: Web, Search, Video, Google, Blinkx
Check out blinkx.com. Blinkx has indexed the audio of "seven million hours of video" found on the web to make them internally searchable. CacheLogic, Cambridge, England, reports that with the YouTube phenomenon, etc. 60 percent of web traffic is now in video heading toward 98 percent. Google conquered the "first, text-based era of the Web," but relies upon often incomplete and inaccurate metadata surrounding videos to render them searchable.  Blinkx's automated speech recognition - "effective speech recognition is a 'nontrivial' problem in the language of computer scientists" - uses "'hidden Markov models'" for contextual analysis. Author describes effectively searching for SNL "Lazy Sunday" video and compares to Google search. Google is probably not ignoring this issue; TruVeo, Flurl, and ClipBlast are in this space, too. And audio is only one side of video - what about indexing and making video imagery searchable? John R. Smith at IBM's T .J. Watson Research Center's working on experimental search engine for videos called Marvel that includes "visual information." John Battelle, author of The Search about Google, makes the point that search is the gateway to the Internet. Expect more. My own thought: this article neglects the important role social networking plays for folks using the web. Like-interested people connect thru social sites or sites managing links. For example, when I used Blinkx to search Frank Lloyd Wright, I soon found a plethora of Wright videos on YouTube just by linking to one from the search. Search is only the first step, when starting cold.