January 25, 2005
Google Video
The Mountain View geniuses do it again with Google Video, a television show search engine. I haven't read up on the back-end technology, but after playing around with it for a few minutes, it's clear they're doing some sort of speech-to-text and associating stills with the conversions. So when you give it a query, the video stills are relevant in context to when the term occured in the show. I searched "obama" and I got back, among other results, the exact still of the Jeopardy! answer about him, with the corresponding transcript highlighted to when Alex was reading the answer and the contestant replied with the question. You can't actually watch any video via the site, so utility seems limited at the moment. It's still cool, in a way that I think is designed to drum up demand for access to the video contained in your search results. Where that demand goes is anyone's guess right now. Content owners think advocates for sensible copyright restrictions are communists.
Okay, they have figured out how to scrape a video (and presumably audio) feed for human language and put it all in the search relevancy results we've all come to expect. They've only been running their indexer against television feeds since December, but think about it: if it weren't for the stingy copyright, we already have the technology to google any TV show, really any media -- music, movies, audiobooks -- ever. That's pretty fucking amazing. It's not surprising, because all of the pieces to do this have been floating around for a while, and it takes a Google to pull it off, but as I was just sitting here thinking of how we could have truly on-demand entertainment ("oh, what was that one episode of West Wing, with the diary? Let's google it and watch it...," or, "I want to see a movie right now about Vietnam," or "How does that song, Genius of Love, go, again?"), it seemed to warrant mention.
UPDATE (minutes later): Alright, on further examination, it appears they are getting the text from speech via the closed captioning stream, so they've offloaded that heavy-lifting to existing infrastructure. It's still impressive, even if it isn't as sophisticated as it initially seemed :-( Reading the "About" page: who knew?
Okay, they have figured out how to scrape a video (and presumably audio) feed for human language and put it all in the search relevancy results we've all come to expect. They've only been running their indexer against television feeds since December, but think about it: if it weren't for the stingy copyright, we already have the technology to google any TV show, really any media -- music, movies, audiobooks -- ever. That's pretty fucking amazing. It's not surprising, because all of the pieces to do this have been floating around for a while, and it takes a Google to pull it off, but as I was just sitting here thinking of how we could have truly on-demand entertainment ("oh, what was that one episode of West Wing, with the diary? Let's google it and watch it...," or, "I want to see a movie right now about Vietnam," or "How does that song, Genius of Love, go, again?"), it seemed to warrant mention.
UPDATE (minutes later): Alright, on further examination, it appears they are getting the text from speech via the closed captioning stream, so they've offloaded that heavy-lifting to existing infrastructure. It's still impressive, even if it isn't as sophisticated as it initially seemed :-( Reading the "About" page: who knew?
