I’m a political junkie in my spare time, so I have the Sotomayor confirmation hearings playing on C-SPAN as I write today. But while I heard some of Lindsey Graham’s surprisingly honest and heartfelt opening statement, I missed a chunk of it. Alas, I don’t have a DVR, so I couldn’t just hit “rewind.”
First I tried to find it using Twitter Search. That wasn’t much help, as it was mostly filled with people talking about one particular quip made by the senator. I tried OneRiot, but was presented with similar Twitter results and no real-time video or transcripts. Google linked me to Graham’s Wikipedia entry, as well as to the official Senate home page — plus some older videos on YouTube. Bing was similarly useless.
I finally found the video by hunting around on C-SPAN’s site and finding a DVR-esque video stream of the confirmation proceedings that let me go back and find just what I wanted to see — but I had to hunt for it. It wasn’t something a search engine could have found for me. Now that some time has passed, of course, the video has been excerpted, making it much easier to find. That’s not really “real-time,” though. That’s just “recent.”
Good point, and you’re right. Real in-depth, intuitive searching on the major search engines, a real thinking crawler, is not something we yet have. I’d like to see more of a move toward this, myself.
Yahoo results have news on Graham at the hearings. Don’t forget the other engines. :)
Sir the issue you are having is that the technology is *not* here ( yet ) to have machine transcriptions of what was said, and made into searchable semantic mark-up. Something that does not exist cannot be deemed a “failure”.
Your personal definition of the “real time: web made need modifying. Tweet at will and it will be cached, searchable, but that’s text-to-text, not human speech-to-text. Google is close with translating speech to text, as seen in your Google Voice account, but it will be a bit longer.
The larger question would be, why isn’t the stenography being performed inside the room, as Sen Graham speaks, being made publicly available?
The technology exists … not 100% accurate at auto transcription .. but it exists. No one has yet implemented it. Probably an opportunity for someone to introduce vtweet … like vblogs
Anand
You’re a complete brat. Relax. Technology is miracle and you’re a douche for wanting it to serve your every little need.
My exact thoughts, Adam! Not everything said or done can quite possibly be at your searchable disposal AS SOON AS IT HAPPENS. Give it 20 minutes & it just might be.
Thus, “the failure of real-time search.”
Adam, spot on mate. The fact that this post was even considered blogworthy in the first place made me gag a bit.
Dear Jordan,
It sounds like you won’t be happy until the entire information of the world is being fed into your brain in real-time.
You’ve got a Virginia ham under one arm and you’re still crying because you have no bread. Get over yourself, Jordan.
Dear Adam,
You ARE correct! I am indeed a spoiled brat. I am glad you are here to let me know this in REAL TIME!
Sincerely,
Jordan
Jordan –
Good post. Clearly one WILL be able to do this someday (no doubt the NSA can currently do this for every television show broadcast anywhere on earth already)… it is mostly just a matter of all the right technologies getting integrated.
@Todd, I suspect you don’t need speech-to-text to solve this problem at all. There’s a little thing called closed-captioning which is common on all broadcast television (I can’t swear it is being done for these hearings, but I’d bet a fair amount of money that the PBS feed has it) and COULD make this all easy – but I don’t see it being put to use for video search – YET.
I thought some readers may be interested:
This technology was explored some time ago (1995?) by a MIT Media Lab grad called Bradley Horowitz and his company Virage. Brad went on to Yahoo and acquired Flickr for them and ran the Brickhouse innovation group.
http://www.virage.com/
Virage was bought up by Autonomy who also invested in a company called Dremedia, started by a BBC guy Matt Karas which did similar things ( index voice ). So if anyone could make this happen it should be Autonomy.
The one company everyone seems to miss is Nexidia. They started off selling their technology to governments and now have moved into similar space of indexing real time voice streams.
So the takeaway is that it is entirely possible to make video that contains narration searchable. Someone needs to productize it for the consumer.