Video search engines help users sort through clips

— -- The Internet is teeming with so much video that searching through it is becoming one of the biggest challenges on the Web.

Video search engines such as Blinkx and EveryZing are among those racing search giant Google to try to solve the problem. Both use speech-to-text and other technologies to make video clips easier to search and view. There's a lot at stake. The video advertising market is projected to grow to $4.3 billion by 2011, up from $410 million in 2006, researcher eMarketer says.

Search technology can read and analyze text on a website, but the same technology is limited when it comes to video. If a website called "All About Dogs" also offers information about cats, a traditional search engine can figure that out. Not so for video.

Most search engines make educated guesses about the contents of video clips based on the coding used to "tag" or identify them, or by the words other websites use to link to the clips, says Kevin Ryan, global content director for Search Engine Watch, a Web information site.

Google hopes to improve on that. The company has just launched a test of a new video search gadget for its YouTube politician channels. It uses speech-recognition technology to create searchable transcripts of videos. For instance, you can search on "health" to find a clip from John McCain, Barack Obama or many others, and even jump to where those key words appear in the video clips.

Other companies tackling video search:

Blinkx: 'Snackable' video

Video search engine Blinkx can take the audio from a video clip, break it down into the smallest distinguishable sounds in any language, and create a transcript of what is being said.

Blinkx also uses the text that appears on screen, such as scores or times displayed during a sporting event, to identify video.

"Video today is very packaged and very linear," says Blinkx CEO Suranga Chandratillake. "There's no reason it has to be like that. It can be snackable."

Truveo: Crawling the Web

Truveo, a video search engine owned by AOL, focuses on "crawling" the Web for video clips. Crawling is the process search engines use to locate media files across millions of websites. Only when a search engine has found and built a list of video clips can users then search for what they want.

One of Truveo's advantages is that its crawling technology can search for video clips on websites built with advanced Web technologies such as JavaScript and Flash.

While Truveo is also using certain social computing concepts (such as user-generated "favorites" lists) to provide relevant results for video searches, its focus is on building the most comprehensive index of video clips on the Web. "You can't do anything in video search if you don't have the video itself," President Pete Kocks says.

Mefeedia: 'Social discovery'

Some video search engines are eschewing audio and video analysis in the short term over concerns about whether those technologies are ready for prime-time use.

Mefeedia relies on users to help sort through millions of video clips and TV shows. They can create "channels" or playlists of videos that others can watch or contribute to.

If one of your friends watches a clip on Mefeedia and enjoys it, the clip shows up in your account, letting you know that it might be worth viewing.

EveryZing: Any search engine

EveryZing, like Blinkx, analyzes audio from a video clip and turns it into searchable text. The main difference is that EveryZing focuses on professional content that its clients produce, and uses its technology to "wrap" video clips in data that include the whole transcript of the speech track. Other search engines can then search within the dialogue of video clips without having to invest in their own audio and video analysis technologies.

EveryZing chief revenue officer Stephen Baker says that his company's software can create transcripts of most professionally produced news programming at an accuracy level above 90%.

Baker believes that as computer processing power gets cheaper, deep video analysis will eventually go mainstream.

He envisions an investor being able to search for every instance of someone like Bill Gates from video clips across the Web in an instant — even if Bill Gates' name never appears in the title or information associated with the clip.