|
|
The Joys of Dictating
I began using speech recognition (SR) software in earnest three and a half years ago after fracturing my right index finger in a cycling accident. Dragon NaturallySpeaking has made a huge difference in the way I translate, significantly enhancing my output while eliminating the aches and pains I would feel in my fingers, wrists, neck, and eyes after a long day at the keyboard.
One downside to SR is that it requires a fairly powerful PC and a headset or microphone, making it difficult to use away from the office. While it's physically possible to don a headset and fire up your notebook at Starbucks or the local park, it's not something I've ever felt comfortable doing. And as light as my ThinkPad is (under 2 kg), I'd rather not carry it around if I don't have to. Enter the digital voice recorder.
Small Package
As far as I know, relatively few "dictators" use voice recorders more than occasionally. Why not? Limited selection (until recently), high cost, poor recognition accuracy, and the inability to perform on-the-spot corrections and revisions.
But things are changing. SR-ready digital voice recorders are now available from Sony, Matsushita, and Olympus, among others, for around \20,000. The good ones use generic memory cards--CompactFlash, SmartMedia, or MemoryStick--which are cheap and provide effectively unlimited capacity. The 32 MB SmartMedia card that comes with the DM-1 holds a little over five hours of high-quality recording time, which in my experience is enough for about 60 pages of dictation. An extra 64 MB card can be picked up for a little over \3,000 and should be sufficient for the most demanding jobs unless you plan to be away from your PC for weeks at a time.
The DM-1 weighs 86 grams with two AA batteries, which is good for about eight hours of use. An AC adapter is sold separately. The recorder fits easily in a front shirt pocket and can be held in one hand indefinitely with no strain. Record and pause controls use the same large, centrally located button. A padded soft case is also included, with a belt loop for the uber-geek. The unit features jacks for an external earphone and microphone (more on this later) along with switches for music/voice mode, microphone sensitivity, and other standard recorder controls. Additional features include stereo output, MP3/WMA playback (software included), and the ability to serve as a SmartMedia reader and writer for your PC. Compatibility with the Mac OS appears to be limited. (In any case, serious users of SR would be much better off with a Windows 2000 system, in my humble, I've-used-both opinion.)
For speech files to be transcribed with SR software, they must first be transferred from the DM-1 to a PC using the included USB cable. If, like me, you use NaturallySpeaking rather than ViaVoice, you must also convert the downloaded file from the native DSS format to a WAV file. This is accomplished with a single right-click.
Microphone Critical
How accurate is the resulting transcription? Using the built-in microphone in a quiet environment, accuracy is quite serviceable as long as you enunciate clearly. In a crowded coffee shop or airplane, though, forget it. Your voice will be lost in a sea of extraneous noise.
Fortunately, noise can be muted and accuracy enhanced by adding an external microphone with a noise canceling circuit. Plugging in an old Andrea headset produced a huge improvement in sound quality and a noticeable gain in recognition accuracy (see test at end of article). But having to wear a headset defeats the purpose of owning a lightweight, hand-held translation device. So I've ordered a tiny lollipop-shaped microphone, the Buddy S35, designed especially for digital recorders. The cordless Buddy plugs directly into the microphone jack and provides the same noise cancellation benefits as a headset, or so the manufacturer claims.
If the Buddy works as promised, I expect to be doing a lot more translation outdoors. Interestingly, I find myself using the DM-1 a lot at home--despite the availability of both desktop and notebook PCs--while sitting at the kitchen table with a cup of coffee, say, or kicking back in a lounger with a beer. For airplanes, coffee shops, and other noisy, public environments, a notebook will continue to be my weapon of choice, although I'm flirting with the idea of buying a handheld device, like the Psion 5mx or AlphaSmart keyboard recommended by Richard Sadowsky, for such situations.
Cons and Pros
As Pai Hwong has noted, the process of dictating to a recorder is linear by nature and does not allow for backtracking and rewriting. Unless you're the type that gets it right the first time, every time, this is going to compromise your work style. Another disadvantage is the inability to "train" specific words. (Note that this may not be the case with ViaVoice.) Naturally, you can still perform on-screen corrections after transcribing the speech file.
Another caveat: handheld dictation probably isn't a viable option if you need frequent access to dictionaries. I work in a fairly narrow field (banking and macroeconomics) and seldom refer to kanji, general-purpose "kokugo," or Japanese-English dictionaries. Proper nouns and abstruse financial concepts are handled using a personal glossary, print references, and web searches on a second pass through the document, following downloading and transcription.
This is not a technology I could use all the time, but it offers three solid benefits. First, being able to get away from the PC for an hour here and an hour there makes a noticeable difference in my alertness and efficiency throughout the workday. Second, it's an ideal tool for use outdoors or for when you're going out and may have time to knock off a few pages but don't want to pack a heavy, expensive notebook. The other day I took the DM-1 on a hike and did some work on top of a mountain. At 86 grams you don't have to think twice about bringing it along. Third, it's a neat tool for dictating letters, notes to yourself, and voice files to send to family and friends. Those with MP3 or WMA libraries can also use it as a solid-state Walkman.
Just Do It
Regardless of whether the idea of PC-free translation lights your fire, I highly recommend giving SR software a go. Both Dragon NaturallySpeaking and IBM's ViaVoice have proven technology and a large user base. Introductory packages can be purchased for around \10,000 with microphone, and more full-featured "professional" versions for about twice that. A fellow translator and DNS user recently told me he would pay $1,000 for his SR software, and I would gladly cough up several times that (long live the consumer surplus!). This is amazing technology, folks.
The Bottom Line: Dictation Test
I dictated a short passage using three configurations: 1) DM-1 with built-in microphone, 2) DM-1 with headset, and 3) desktop setup with headset. Conditions were optimal: I read from a script, which naturally results in higher accuracy than dictating on-the-fly, and I took pains to enunciate clearly, which is easy to forget after three or four hours of translating. Here are the results, with errors underlined [here in red] and omissions in brackets:
Chris Green lives in Tokyo. When not translating reports on banking and the economy, he enjoys reading financial history and pondering the ideal gear cluster for his touring bicycle.