With today's overwhelming popularity of Web videos and the many sources that make them available, the biggest challenge now is to know exactly what to look for and where to look to find it. For most of the Web video sites like YouTube, their contents are catalogued by the their original publishers using tags and keywords or "contextual metadata" before they are uploaded for the search engines to be able to find them. One of the drawbacks of that practice is the results can be very unpredictable as they depend on the accuracy of the tagging itself. And since there was no way for a human or a machine to double checking the tags and keywords used, the potential for abuse and misleading were very high. Now, a few start-ups have seized on the need for a more reliable way of sifting through millions of videos and pick the right one. One of those start-ups is a California based company called VideoSurf Inc. What distinguish this video-search site from the rest, is its revolutionary concept to conduct search by "seeing" through a video's images for a particular content as well as using metadata. The company claims to have analyzed and categorized more than 12 billion visual moments lifted from various Web sources at its site VideoSurf.com. Another unique characteristic of the site, is the way it displays the search results in a film stripe like format differentiating one scene from another. It also allow users to search by showing only people faces' , useful if you are looking for a particular person in a movie or a long video. The site is in its public beta phase which may explains its lack of a hundred percent accuracy for displaying the right video or listing the actual available ones. It plans to be much improved and fully operational early next year.
For more, see WSJ.com
No comments:
Post a Comment