How do I turn recorded speech into text?

March 20, 2006 3:19 AM

What is the easiest way to turn a recorded interview into text?

I have recorded an interview onto my iPod using Griffin iTalk. What is the easiest way for me to now to transcribe this into text? It doesn't matter too much if it is not a very good conversion as I have to go through it all anyway but I'm looking to avoid typing the whole thing out from scratch. I've got a desktop PC and an iBook.

Thanks

posted by janecr to computers & internet (15 comments total)

You can use a transcription service - email the mp3 to them and they will return a typed transcript to you. Cost dependent on number of words I think. I've never used one of these services so I cannot recommend one.

Voice recognition software just won't work in my experience, if that is what you are hoping for
posted by ajbattrick at 3:24 AM on March 20, 2006

I agree that good, old-fashioned ears-to-keys is really the best method here. I can't imagine there's a good enough version of (for lack of a better term) "audio OCR" that could work well enough with dialects, variable recording qualities, and unique speech signatures to produce a quality interpretation.

Also, if you'll be beating this text into a reprintable state later, it helps to have the whole context of the transcribed interview in your head if you need to re-interpret an interviewee's line for clarity or rhetorical sharpening. A flat "ah um" on print means a lot less than the sound of a particular "ah um" on tape, if you know what I mean.

Then again, I may just be a luddite, and some fabulous new software is lurking in the answer just below mine.
posted by mykescipark at 3:35 AM on March 20, 2006

If you don't find an automatic solution, Express Scribe is a good program for manually transcribing stuff. In the preferences set up keys you don't use much (like square brackets and backslash) to be playback controls, allowing you to manually pause and rewind without taking your hands off the keyboard.
posted by cillit bang at 3:52 AM on March 20, 2006

Transcription services are great. If you can't spring for that, find someone you know who can touch type, as if you can't touch type trasncribing audio is horrifically slow. If you can, it's much faster than you'd think.
posted by unSane at 4:35 AM on March 20, 2006

I'm going to go with typing it. If there was effective cheap software for this then I wouldn't know a half dozen undegrads who get paid well above min wage to type up interviews and transcribe lectures.
posted by tiamat at 5:15 AM on March 20, 2006

Similar questions have been asked several times. As far as I know, all existing solutions require extensive training - you could make your interview subjects read the training paragraphs :P

Given the negative answer, I don't have the will to search out the old threads.. I'm pretty sure I haven't seen Express Scribe recommended yet though, looks good!
posted by Chuckles at 6:18 AM on March 20, 2006

Maybe ScanSoft Dragon Naturally Speaking 8 would work? saw it used on amazon for 75, might be cheaper elsewhere. havent accually fiddled with it meself.
posted by psychobum at 6:24 AM on March 20, 2006

I know I've head about a service where you upload a file and get back a transcription.

It made the rounds as the site-du-jour on some podcasts I listen to. I definately heard about it within the past two weeks.

Darned if I can find the sight now though. Argh.

googling for 'podcast transcription' might provide some interesting leads.
posted by Wild_Eep at 7:29 AM on March 20, 2006

I've head about a service where you upload a file and get back a transcription

Many (most?) transcription services now offer that option. Some of my friends work at this one, which has offices in New York and Raleigh and has a link on the home page to upload digital files to a secure server. If you decide to do it by hand, definitely get a program like Express Scribe; it will save loads of time.

Side question: Has anyone invented a foot pedal control that interacts with digital media? Using a foot pedal was great for transcribing tapes.
posted by mediareport at 7:55 AM on March 20, 2006

Side question: Has anyone invented a foot pedal control that interacts with digital media? Using a foot pedal was great for transcribing tapes.

Good question. Maybe you should open a new AskMe thread for this, because I would be very interested in this.
posted by NekulturnY at 8:29 AM on March 20, 2006

The Express Scribe page has a link to purchase a footpedal. I'm assuming it only works with that particular software but it could be pretty damn useful.
posted by vetiver at 8:53 AM on March 20, 2006

Dragon requires a training curve for both the software and the user. It is also stongly sensitive to audio quality, voice speed, stress and inflection (I could always tell when it was time to take a break because my accuracy went down due to voice stress). Now I've transcribed interviews by listening and speaking into Dragon, but I don't know if it is that much better than what you can do by hand.
posted by KirkJobSluder at 9:42 AM on March 20, 2006

I know someone who did it with Dragon Soft -- but he had to first sync the software with his voice, and then repeat what the interviewees said.

I touch-typed mine. It took about 4-6 hours/hour of tape, I believe (this was 2 years ago, so my memory may be off).

(There does exist some sort of keyboard triggered program that will pause a tape and then return to playing it, but I didn't use it because my tapes were manual.)

One trick that helped with this -- in Microsoft Word, Tools->AutoCorrect -- you know how if you accidentally type "teh," it changes to "the?" I programmed tons of common words and got used to these keyboard shortcuts. "J" expands to "and," "p" to "people," "ddn" = "didn't," and so on. This was especially helpful for people's verbal ticks ("idk" to "I don't know") and for long words related to the subject matter ("neighborhoods" and "communities.")

Good luck!
posted by salvia at 6:07 PM on March 20, 2006

The Express Scribe page has a link to purchase a footpedal...it could be pretty damn useful.

Foot pedals are *extremely* useful for transcribing tapes, and I imagine the benefits are just as real when transcribing digital media. Setting keyboard controls is nice, but a foot pedal allows you, for example, to keep typing while rewinding to check something. Saves lots of time once you get used to the rhythm of the thing.
posted by mediareport at 11:13 PM on March 20, 2006

Thanks, everyone, I downloaded Express Scribe and can recommend - is making the whole thing much easier...
posted by janecr at 4:42 AM on March 27, 2006