Darpa Wants You to Transcribe, and Instantly Recall, All of Your Conversations | Hacker News Reader

Darpa Wants You to Transcribe, and Instantly Recall, All of Your Conversations

wired.com

16 pointssk2code11 comments

You know what a great use for this system would be? Taking all those videos of talks, video tutorials, and so on, and making transcripts. I think it is cool that there are now howto videos for everything, and most talks worth watching are now online - however somewhere along the way text got lost. It sure would be nice to not have to watch a whole 30 minute video when I could skim the transcript for the bits that I want to see (timestamped) and jump the video that point. Even more useful would be auto-annotating slides from the talk recording. It's fun browsing slides, but more fun would be to read the talk that went with it.

Google could probably implement this with Youtube. Their automated translations are a start, but if viewers could edit the transcripts as the video plays, they could come up with good transcriptions very quickly.

A UK expert of sorts covered this very well. Watch [0] or read a spoiler full review [1] or google for more reviews

http://www.channel4.com/programmes/black-mirror/episode-guid... http://blogs.independent.co.uk/2011/12/19/review-of-black-mi...

This episode was optioned by Robert Downey Jr. to be turned into a film.

It occurred to me that just as type setting has caused some standardization in written letters, and the written form of language has caused standardization in spelling, that the necessity to speak clearly to a computer could cause a similar standardization in verbal communication. It would be interesting if we adapt more to the machine than the machine to us, learning to be a little more careful about enunciating and learning to use commands in speech that are meant for the computer, like saying "comma" in a sentence.

Interesting point, as I often find myself (being a human) having a difficult time understanding many people, mostly due to their unfriendly communication patterns (low talking, mumbling, not enunciating, drawls/accents, etc).

Seems like one of those things the general public pushes back on, if for no other reason than virtually everyone gets offended when told they don't know how to speak.

I use this one a lot: "I can't understand what you're saying."

I figure it's closer to showing someone "your attempt at communication has failed" versus telling them "you don't know how to speak".

YMMV.

What I really could use today, is software at my computer that records my whole workday in front of it, and allows me to go just to the points where I was making sound later on, and easily extract worthy material.

I explan: I often spontaneously start reciting some poetry or prose come from god knows where in my subconscious, in the form and meaning of a speech, a pitch, a story or even a song. Sometimes it's just a beat or a melody. These often turn out to be incredibly creative and surprise me, but by the time I setup something to try and record it, surely as winter comes every year the creativity has fled the building, and I can't record anything worthwhile. My creativity is like a lovely squirrel who stops playing with the acorn once he notices he's being watched.

I have looked into surveillance software now and then after such an episode, but really, this should be easier. This should be an always on Evernote audio note with soundwave analysis strapped on. I wonder what artists and even writers use in their studios and studies, for surely those who don't have such a setup are missing out on a lot of their greatest, most spontaneous material if they're anything as eccentric as myself.

You wouldn't even have to keep recordings of the whole day. I learned a neat trick from watching the making of the show "Planet Earth". To record some of the amazing unexpected events, they would have a camera running constantly that keeps about 10 seconds of HD video in memory at a time, then deletes it. When they see something cool happen, like a shark jumping out of the water, they just hit a button. That button tells the camera to not delete video that is more than 10 seconds old.

That way they didn't fill up their hard drives and they got the unexpected stuff.

What you could do is have a sound analyzer listen for speech and it would be like pressing that button, you'd get whatever happened 10 seconds or whatever, before you started spontaneously jabbering. :)

This is sound advice, pardon the pun. Still I wonder if I should ramp up the TiVoish recorder to about an hour or so, because often these are so spontaneous in fact, almost subconsciously uttered, that it might take me more than 10 seconds to even realize I'm doing something weird, let alone to figure out I'm on to something.

Given storage sizes these days you could have that buffer be a full on hour than you can grab from.

Reply on news.ycombinator.com