"Interesting stuff of the Technology, Products and Web 2.0..."

Wednesday, September 17, 2008

Google Audio Indexing aka GAudi Goes Live

Nearly two months ago, Google introduced the Google Elections Video Search gadget on iGoogle, a tool that transcribes and indexes the spoken content on YouTube's Politicians channels.

Now it's possible to enjoy this technology in a bigger way: Google just launched Google Audio Indexing (aka GAudi) in Google Labs. The dedicated site offers more features, such as "search within video" and "sharing," and a more robust user interface.

Google Audio Indexing is a new technology from Google that allows users to better search and watch videos from various YouTube channels. It uses speech technology to find spoken words inside videos and lets the user jump to the right portion of the video where these words are spoken.

Both Google Audio Indexing and Google Elections Video Search gadget, are using the exact same underlying technology. The new features from Google Audio Indexing are being incorporated into the gadget.

The scope of the Google Elections Video Search gadget is restricted to the US election. The aim of Google Audio Indexing on Google Labs is broader and the US election is just a first step. With Google Audio Indexing, you can use speech technology to find videos and to search for spoken content inside a video.

Searching for videos

To find a set of videos, simply type a query in the search box and press the "Search videos" button. The results of your search will appear in the left column. You can refine your search using channel filters, the active filter showing in bold. A channel filter corresponds to one or more YouTube channels. In the context of the US election, you can choose videos from the McCain channel, the Obama channel or from all YouTube political channels. For each result, it provides a thumbnail of the video, its title, the time since it was published, the duration and the number of times the query terms are mentioned (i.e. spoken) in the video. By clicking on a result, the video will be displayed in the right column. If you need to see more results, you can navigate between result pages using the links at the bottom of the left column (previous, next, etc.), just like on most Google Search products.

Watching a video

Once you have selected a video, it will appear on the right side of the screen, in the embedded YouTube player, with the list of mentions. Google Audio Indexing enhances the YouTube player to display mentions of the query terms as yellow markers on the player timeline. For space reasons, it cannot always display all the mentions on the timeline. The top 10 mentions are displayed as a list under the player.

To read the transcript for a given mention, just mouse over the corresponding yellow marker. To listen to the corresponding audio, just click on the marker: the player will automatically jump to the right time in the video. You can also click on the play button located on the left side of the mention.

Searching inside a video

With Google Audio Indexing, you can also restrict your search to the content of a given video. By default, the "Search inside this video" uses the same query term as the video search.

To search for a different term, just type your query in the text form under the video player and press the "Search within this video" button. The mention results are displayed underneath. The yellow markers on the timeline are updated accordingly and you can access the information just like before, using mouse over.

Sharing the video with your friends

You can share a given video with your friend. Just click the share button and copy-paste the URL. When your friend receives and clicks the URL, she will be redirected to the same page, the same query and the same video displayed. Note that the search will show a single result, the video being shared.

Google Audio Indexing uses speech technology to transform spoken words into text and leverages the Google indexing technology to return the best results to the user.

The returned videos are ranked based -- among other things -- on the spoken content, the metadata, the freshness.

It periodically crawls the YouTube political channels for new content. As soon as a new video is uploaded to YouTube, it is processed and then available in index for people to search.

No comments: