In October I asked your advice about using a digital voice recorder for transcription, primarily to smooth out my reading workflow (see How to read a lot of books in a short time). I bought one and I've been using it since mid-November. Following is a report on my experiences and recommendations.
WS-300M: A nice product
I opted for the Olympus WS-300M, a combined voice recorder and music player. It's about $80 on Amazon, and is compatible with Dragon Naturally Speaking (I took Eric's advice to check hardware compatibility).
The unit itself is great, though a tad pricey and pretty small - so I have to be careful not loose it. Main likes:
- USB mass storage - Just plug it in and voice files (WMA format) are immediately available. No drivers, etc.
- Music player - I admit I haven't used this, but I like the option. Not sure the output amp has enough power...
- Nice selection of features - record quality, mic sensitivity, folders, etc.
- Recording time - Long! I use the HQ mode, which provides over 17 hours. (You can get 66 hours in LP mode.)
There's really not much to dislike. The built-in mic has a lot of handling noise (understandable, and you can buy an external mic), and having only two mic sensitivity settings is a bit all or nothing, but overall it's not holding my reading workflow back.
Sadly, the same isn't true for Dragon.
Dragon: Not so good
The program is pretty good in some ways, but I had a number of issues with it, primarily:
- Recognition rate too low - I had to make a lot of corrections.
- Making corrections was too slow - The program learns via corrections, so this is important. However, waiting for the menu to pop up was excruciating.
- Slow overall performance - Start up, recognition, etc. all took a long time. Clearly one needs a very modern machine to run this on.
- Poor quality headset - The included headset broke after only a few uses. Official replacement price: $30.
Ultimately, performance was a deal breaker, and I've concluded that manually transcribing and typing is faster.
As an alternative, you can have someone else transcribe for you. Some possibilities (I haven't tried them): CastingWords: Podcast Transcription Service and SpeakWrite: Voice to Document Dictation Service.
(You can find more on typing vs. recognition on the Davidco forums: Typing at 300-400 words per minute?, Best Voice-Recognition Software??, voice recognition, Voice Recognition Software, Help! Touch typing software?, and Alternative to Mavis Beacon's Typing?)
So how does it all come together? It's simple, really. I record notes as I'm reading into the recorder, usually during exercise or those "between" moments (waiting for an appointment, etc.) Longer reads are broken up into separate recordings, one per session. The recorder has five folders (each capable of holding tons of files), and I use one folder for each book/video/podcast I'm absorbing. Once I finish, I create a "transcribe ____" action.
When I get to the transcribing, I fire up my editor (see My Big-Arse Text File - a Poor Man's Wiki+Blog+PIM), download the audio file, and start listening and typing. To make this efficient, I never use the mouse. Instead, I switch rapidly between the two programs (alt-tab), using the pause/play shortcut (ctl-p) to catch up on the typing. Simple!
This approach works great, and has led to a few surprises.
First, I like having the five folder limit - It keeps my eyes from being bigger than my stomach. Otherwise, I tend to keep reading (i.e., creating more work), and avoid the (more difficult) transcribing part. However, once they're full, I "heat up" those transcribe actions to make some room.
Second, because the recorder's so small, it's easy to have with me, so I'm not having to capture reading notes in multiple places - it's all in the recorder.
Third, I used to abhor transcribing - it was one of those actions that seemed bigger than it was. However, I've trained myself to look at the play times for the files, and they're just not that long! I'm more likely now to just jump in and do it.
Finally, I've found I really want the recorder with me all the time. Not just for reading, but because I'm noticing sounds I want to share with others, especially my six-year-old's delightful impromptu singing, which is really precious to me.
For the future, I need to improve my typing speed and accuracy.
Also, I'm experimenting with ways to make sure I use what I learn from the reading. Specifically I've created a script that nightly selects a random entry from my text file and emails it to me - a kind of "brain blast from the past." More later.