Google Audio Indexing: Keyword Searches for Political Videos
Two months ago, Google released a widget for iGoogle that allowed users to search for keywords in election related speeches on YouTube. Today, Google expanded the scope of this search to a larger selection of political videos and the widget graduated to become a full-blown Google Labs product named GAUDI. Interestingly, the tagline in Google Labs describes the product as a search for "what people are saying inside YouTube videos," even though its focus so far is very narrow. Over time, we expect Google to expand this service to cover a larger selection of videos or even all of YouTube.
Only Politics
Google currently only indexes videos in YouTube's "Politicians" channel, which includes both speeches and political advertisements. As Google explains, political speeches are a relatively easy target for speech recognition software because they have received a lot of academic and industry attention and are known to perform well. However, in the GAUDI faq, Google also hints at the fact that Google expects to expand this product beyond these political speeches in the future.
Great Start, But Interface Needs Improvements
While the interface is very well designed, we can image a number of improvements. It would be extremely useful, for example, to be able to sort videos by date. Right now, they are always sorted by relevance, though in this fast-paced political season, it is often more important to know when somebody last used a certain phrase.
Also, while the variation on the YouTube video player Google uses here is well suited for this task, you can not embed the videos on your own site. All you get is a link to the video on the Audio Indexing site.
As of now, Google also doesn't offer an API for this service. Given Google's record, however, we assume that it will make an API available in the near future.
At times, like all speech recognition systems, GAUDI will also make mistakes. Overall, however, we came away impressed with the quality of the transcriptions.
Share
Comments
Subscribe to comments for this post OR Subscribe to comments for all ReadWriteWeb posts
They cannot release the API soon enough!
I suspect that this is really proof that Google and everyone else's attempts at monetizing video isn't working ( contents of video cannot be indexed in search ---> no AdSense revenue ).
My first impulse when I read about audio indexing was privacy. Do you really want everything you have ever said in a web video searchable via Gogole?
We take care in what we write, always being professional and polite, but I am sure there's a YouTube video or two out there where I am drunk, using the f-word as a noun, pronoun and an adjective - all in one sentence.
google.com this experimenting with all this right now
They are certainly testing this on various of their own products.