Skip to main content

Before multichannel television, viewers selected programs by channel number. In the early days of television, that may have been the VHF or UHF channel number. Now that channels form part of a multiplex, the concept of selecting RF channel ceases to have a meaning to for the viewer. And as the number of channels broadcast increased, it became difficult for the viewer to remember the channel number, but the EPG provided the answer. Coupled with the advance of channel branding, this concept works well for up to around 1000 channels.

I recently attended the annual summit of the Digital Television Group, an industry alliance representing broadcasters and CE manufacturers that promotes standards for DTV in the UK. One of the day’s topics was the Internet-enabled TV, which is a receiver that can access not just IPTV services, but any video content on the Web.

Just as the EPG has replaced the basic remote control with channel change and volume control, what is going to replace the EPG to find video content on the Web? To find information on the Web, we use a search engine. One of the speakers, Suranga Chandratillake of Blinkx, gave a presentation on video search. I have a special interest in this, having researched the subject while writing about digital asset management.

The audience learned that because conventional search engines are text-based, they can only glean metadata from TV listings and closed captions. A true TV search engine must use video and audio analysis.

I got to thinking, how do we find what to watch? Conventional television publishes schedules as listings and program guides, and employs onscreen promotions. But beyond the walled garden of conventional TV — out in the world of Internet video — viewers find content by many other ways, including by accident.

One way is via the viral marketing of social networking. To me, this only a small step beyond recommendations at the water cooler. What I want to know is how I can discover content, without peer recommendation or search? I can only search for what I know. How do I find what I don’t know about, but may find interesting?

I know of two automated approaches. One is the search engine; the other is the recommendation engine. These can, of course, be used together. Both these technologies are complex, and to achieve highly relevant results, they must use sophisticated algorithms. I can think of some bad examples on retail Web sites.

The popular search engines are text-based, which is not good for video. Textual information may include closed captions and TV listing, but gleaning the real information needs video and speech analysis, nontrivial technologies if you want the high performance necessary to index thousands of hours of content. The more sophisticated search engines use techniques like contextual search to try to make sense of metadata the way humans do.

The content recommendation engine is also a real product, so there is technology available to supplant the EPG. As governments encourage the spread of wideband IP networks, the viewing of video content from the Web will increase, as will the video quality. To navigate this global content repository, in the future, the EPG just won’t cut it.

Send comments to: