User:MishnayicHacker
From TheCommandLineWiki
Listener and fan of TCLP.
Professional transcriber and editor. Before the recession, I built and managed the team that produced transcripts for NPR. (These are available for free download on each news story's page.)
Before that, I taught French in university while pursuing a few degrees in that field.
As an undergrad I majored in CS but didn't stick with it. Now I am renewing my love of hacking with TCLP.
As a day job, I sell backpacks in Jerusalem.
FAQ
1. What setup do you use for transcription?
On a PC platform, I use the free Express Scribe software from NCH, an Australian firm. I highly recommend this software, which is supposedly available also for linux and OS X at that link. For the record, I am not affiliated whatsoever with NCH or with the coders of this software, but all of my transcription employees appeared to be happy with the choice and never asked to use an alternative. In fact, some have gone on to other transcription jobs where a different setup is used, and have expressed regret that they can't use Express Scribe.
As it happens, I have not been successful running it under OS X. So on a Mac I simply play the audio with QuickTime and pause using the space bar. It's a more simplistic solution to Express Scribe, which can play audio at different speeds and rewind and fast-forward with ease, and which has customisable hotkeys. But QuickTime is doing just fine for the job now, and I don't feel any lack in efficiency.
It's worthwhile to note here that it's important to do the job in two passes: one brute transcription pass and a second proofreading pass. The transcription pass involves typing chunks at a time, pausing the audio in between (although others choose to play the audio at a slow speed, one with which they can type without pausing). Then go back and listen to the audio again while going through the transcribed text with a fine-toothed comb. Something always comes clearer the second time you hear the audio in its full context. This is also a good opportunity to research the spelling of proper nouns and find outside links where applicable. I would recommend not doing that in the first pass, since it tends to break up the flow of efficient transcription.
It's also important to make backups, since even small paragraphs represent several minutes worth of work. But you probably know that.
2. Why don't you just use speech recognition software?
In short, that technology is not ready yet. I look forward to the day when it will overtake a human's ability to decipher every word and to insert correct punctuation, and in fact that will topple the transcription outsourcing industry as it's set up today. But we're not there yet.
To the best of my knowledge the commercial speech recognition software that is the leader in the field is Dragon Naturally Speaking, and I have tried to use it many times (in combination with NoBrainer Univoice). Naturally Speaking uses the same engine as Dragon Dictate for the Mac (both from Nuance). I am sure that these software packages are effective when used for their intended purposes, and this is not to discount them in any way. I am simply stating that the transcriptions that I have produced using software-only approaches (even with a variety of manipulations of the audio) need far too much human editing to be worth the trouble. Nothing I have seen yet matches the efficiency of human work.
There is also an open-source package called Sphinx (http://cmusphinx.sourceforge.net/sphinx4/), which I must admit I have not tried to use yet. The complexity of learning how to install it and implement it is holding me back at this point. I am no longer transcribing for business purposes, so it is difficult to justify the time investment.
The closest, most accurate automatic speech recognition that I've seen is that of Google Voice, but unfortunately I have not had the opportunity to test it because the service isn't available outside the US. I am eager to run some trials and to see if it is already a game changer. If anybody will conquer this field this soon, it's Google.
If anyone is interested in experimenting with these or other methods, I'll be happy to collaborate.
3. How long does transcription take?
4. Should there be one or two spaces after a period?
While Grammar Girl reported once that this was a burning issue in the proscriptive English grammar world (although it's a matter of style, not grammar), the short answer is that if you are writing in html it doesn't mater. Take a look at the following comparison.
This sentence is followed by one. And this is the next sentence.
This sentence is followed by two. And this is the next sentence.
Of course, if it is typeset in a monospaced font, that difference becomes clear. But in proportional fonts, html makes a double space look like a single, so the typist's habits or beliefs in this holy war become inconsequential.
4. Should punctuation be inside or outside of quotation marks?
This is another issue of style. Please refer to the Jargon File, Chapter 5, beginning with the second paragraph.
5. What does the "Mishnayic" mean?
6. With your diverse background, education, and skills, you sound like just the guy we need to hire.
Indeed. Please contact me through my Google profile.
