The video search gold rush

Michael Rogers Columnist • Profile • E-mail

Over half of all online users have already watched video on the Web, and it’s still early days, as cable channels and television networks, along with millions of individuals, rush to upload vast quantities of content onto literally thousands of different Web sites. So the question arises: how do you find what you want in the midst of this utterly disorganized video bonanza?  Clearly, the entrepreneur who comes up with the video equivalent of Google would be in a very powerful position. Digital pundits put the question this way: Who will be the TV Guide of the Internet?

The TV Guide goal is ambitious: during the magazine’s heyday in the 70s it was the largest circulation magazine in the world. Back then TV Guide seemed to be on every coffee table in the country, and it made its founder, Walter Annenberg, a billionaire who went on to be knighted in Great Britain.  Not bad for a magazine whose primary reason to exist was describing where and when to find “The Brady Bunch” — which, in an era of only three television networks, was not exactly rocket science.

As cable and then satellite choices began to proliferate in the 80s, TV Guide struggled to keep up with a myriad of increasingly-complex local editions — but ultimately diversified into the “interactive programming guides” that are now provided in the on-screen navigation for most cable and satellite systems.

While TV Guide magazine continues, with less than a sixth of its 70s circulation, the bulk of the company’s revenue now comes from various versions of the on-screen guides (which now reach far more people than the magazine ever did). Now, of course, arrives the next challenge: a world of Internet video that dwarfs even the 500 channel universe of cable and satellite. Next month, TV Guide plans to take that on as well.

It’s good timing: on the Internet, video search leadership is still up for grabs. That’s because searching video is fundamentally different than searching text: the content is sound and images, and there’s no simple way to automatically search that combination. At present, the easiest solution is to look at the text description attached to the video file by its creator — its so-called metadata. But the quality and format of metadata can vary, and sometimes video clips carry none at all. The next step up is to have viewers themselves “tag” videos once they’ve watched them. But user-generated tags can be unreliable. They require constant and diligent user participation and the quality may vary wildly — starting with elements as simple as spelling (uh, is it “Hilton” or “Hillton”?).

Newer video search companies have come up with some interesting alternatives. Blinkx, one of the current leaders, uses automatic speech recognition to “listen” to each video’s soundtrack and pull out key words that it can search — a notion that was initially used some years ago to search television news videos by “reading” their closed captioning information. A competitor, Truveo (now owned by AOL), looks at all the content on the page surrounding the video to get a better sense of what the video itself is about. And there are a dozen other start-ups in various stages working on variations of those approaches, such as looking at which other Web pages point to a video.

Out on the horizon, researchers are working on teaching computers how to actually recognize images and perform what’s called “scene summarization,” so that future search engines can both watch and listen to videos to determine the content. But that’s tricky indeed.  The first implementations will probably be limited to some subset of images: celebrity faces, for example (thus eliminating the need to know how to spell “Hilton”) or perhaps specific merchandise (“Show me coffee tables that look like this one”).

Regardless of which site you use, current video search is still hit-and-miss: results tend to include videos of radically varying content and quality. Search for “American Idol” and you may well come up with several dozen amateur “American Idol” satires before you actually get to a real clip from the show. (And then some of those actual clips may turn out to be amateur uploads onto YouTube, of dubious video and audio quality.) In other cases (such as Academy Awards footage a few weeks ago) the links may not work at all: after they were posted, the content owners insisted that the clips be taken down for copyright reasons.

MORE FROM PRACTICAL FUTURIST

Practical Futurist Section Front

Will the Internet transform storytelling? What evolutionary psychology says about social networking At this radio network, the audience is the star Is the Internet dumbing us down? Who will fix my wired (and wireless) home? Will road warriors choose electrons over airplanes? Who will be the TV Guide of the Internet? Why is U.S. always last for new phones? Weighing the risks of biotechnology What will make the laptop obsolete? Practical Futurist Section Front   Add Practical Futurist headlines to your news reader: