Google's speech-to-text technology is huge search advance

[hari]

A new speech-to-text technology from Google in which you type in a word or a phrase and it searches through YouTube videos to find all the instances where someone said it.

Right now, it's being used on YouTube's You Choose video page and applied to all of John McCain and Barack Obama's YouTube videos, both those produced by their campaigns and by everyday folk. Look on the lower left hand in the search box and type in whatever you want to hear them talk about, be it global warming, the surge in Iraq, abortion, gay rights, whatever.

If you want to see a transcription, as the video plays, put the cursor on the timeline and you can read along, too.

For now, this new technology is being used only for political videos. But you can be sure it will expand. And you get snippets, not necessarily the entire clip or soundbite, so getting full context can become an issue. But you can easily identify the event or video and then search on it and then watch it in its entirety.

Google does all this through automated spiders, Web-crawling tools that are able to process and transcribe words, which are then embedded as metadata into the video itself.

That metadata — information that describes the context, content and structure of files and documents — can be indexed and searched, just as text.

Google calls it the Elections Video Search Gadget.

"Using the gadget you can search not only the titles and descriptions of the videos, but also their spoken content. Additionally, since speech recognition tells us exactly when words are spoken in the video, you can jump right to the most relevant parts of the videos you find," according to a post on the official Google blog.

Google researchers concede that some of the transcribed parts of the gadget "may not be 100% accurate." But by quickly bringing up the actual words and video segments, users can follow along and fill in any missing blanks.

Google may be the biggest outfit out of the box with this, but two smaller but very competitive rivals, Blinkx and EveryZing also use speech-to-text technologies to make video content easier to search and watch.

And Adobe is also deep into this technology, close to providing similar speech-to-text search capabilities for its Flash, one of the most popular options for displaying Internet videos.

All these rushed and improved technological search advances come as the Web becomes more and more of a distribution center for video.

Online video advertising is booming and expected to grow from $500 million now to $4.3 billion a year by 2011, according to the eMarketer research firm and U.S. Census Bureau estimates.