Susan works in my office. She has to go to meetings and transcribe the minutes. I’ve done transcription and it’s a pain in the ass. Like those kids in class trying to scribble down everything
Doing a few minutes of research on it, and this is the best HTML editor I have access to at the moment.
This Panasonic is the ideal solution, but it doesn’t really work (review).
Actually this device is pointless. It records onto a normal data recorder, and then you feed the audio into a PC and Dragon Naturally Speaking will perform Voice Recognition on it. So the real issue is the software. Here I quote from the review.
“Once you’ve trained the software, you’re ready to transcribe a voice file into text. You can train it to learn several voices independently, but it can’t transcribe a two-way conversation because it can’t differentiate between speakers. The software only works with one voice at a time. “When you record an interview, the software will always try to recognize all speech using the Rebecca user file,” says Matt Revis, product manager at ScanSoft, which sells the Dragon software. “When the interviewee is speaking, the accuracy will be very poor because the software thinks Rebecca is speaking.”
I tried not to take that personally and dictated ” ‘Twas the Night Before Christmas” into the recorder, connected the device to my PC with the included USB cable, and then translated the digital recording into text with the Dragon software. Here’s exactly what came out:
“29 before Christmas Mountain house that a creature was during any amounts stockings around managing you care and hopes the single is to be there until groupware nestle most minivans while visions of sugar plums and their hands and Mom and her cricket and I and my Interest settled down from long winter snap…”
here’s another quote from an informal case study:
We bought Dragon Naturally Speaking to try and automatically transcribe some of the interviews. Forget about it. DNS works GREAT for me on my PC, but it was about 5% accurate with transcribing an interview. It just can’t handle two or more people talking.
One possible variation on the Dragon Model is to treat each speaker as an individual, mike each speaker and hook them up to their own laptop. Then it’d be like they’re dicating to Dragon, comme d’habitude. Then you’d get transcripts of each individuals contribution. And you could time stamp and merge them which’d be a pain in the bum but not an especially complicated bit of Coding. It’s clumsy and I don’t know if it would work, however.
canadalawyers.com claim that they have Voice Recognition that can track 32 different voices and transcribe meeting, but I don’t know.
uh, I think Susan was right. Assuming that Dragon works effectively enough, I could probably cobble together a specific solution for a few meetings … having each person run their own speech recognition off their own laptops, but that even may not work. General Multiple-Voice Recognition doesn’t seem to be market-ready. Windows doesn’t really support or encourage Voice-Rec, even though it’s a great feature, so the market is stagnant. I’m certain that there are high-level corporate or military applications of multiple-voice recognition, but market forces haven’t made them consumer ready yet. I’ll keep looking… but I doubt any of them work well enough.