Google rolls out audio search for YouTube videos

Computerworld - Google Inc. has launched a test of an audio search indexing system that's designed to allow users to find words that are spoken in videos and jump to the portion of the video where the words are used.

The Google Audio Indexing (GAUDI) tool, developed by the company's Google Labs project, uses the same internally developed underlying speech technology as the Elections Video Search Gadget that the company rolled in July. The GAUDI tool will initially be available only for YouTube election videos, though Google said it plans to eventually offer it for use with other videos.

"As more and more video content is being created every day, Google Audio Indexing tries to make it easier for people to find and consume spoken content from videos on the Web," Google noted in an online FAQ. "We see it as an experimental platform where we can learn what features make the best user experience for people looking for spoken content on the Web."

GAUDI works like this: A user types a query into a search box, and then he can refine search results using channel filters, which correspond to one or more YouTube channels. For example, a user could choose videos from the John McCain channel, the Barack Obama channel or from all YouTube political channels.

The search results include a thumbnail of the video, its title, the time since it was published, the duration and the number of times the query terms are spoken in the video, Google said. Users can click on a result to display the video itself. Mentions of the query terms are shown as yellow markers on the YouTube player timeline. Users can mouse over a yellow marker to read the transcript of the words. Clicking on the marker will play the audio.

The technology also lets users query inside a video and share results with other people. Users can send a URL to friends, who can click on it and be redirected to the same page, the same query and the same video from the original search, according to Google.

GAUDI uses speech technology to transform spoken words into text, and then indexes the text using Google's search technology. Google crawls YouTube political channels for new content, and as new videos are uploaded, they are processed and made available to GAUDI for indexing.

Ionut Alex Chin, a blogger at Google Operating System, noted that the GAUDI interface is attractive because it allows users to find all of the mentions of their keywords and jump directly to the appropriate sequence in the video.

However, he also noted that the service does have hiccups. For example, "in the video 'Obama on the 40th Anniversary of the Prague Spring,' the word Czechoslovakia is incorrectly detected as tech also but there," Chin wrote. Moreover, "free is replaced by forty, and there are many other mistakes," he added.

Read more about web 2.0 in Computerworld's Web 2.0 Knowledge Center.