Friday, 31 December 2010
Dictation and the machine
Now I'm adding another tool to the literary workshop. At Christmas, my in-laws bought me the Dragon NaturallySpeaking (sic) software. I can't imagine where they got the idea. I suppose it's because I'm a freelance writer and they figured that I was already well-stocked on thesaurae and dictionaries. I couldn't see much use in it, but I've always been interested in VR technology and I found that being able to order my phone to place calls without taking it out of my pocket was suitably futuristic, so I figured what the heck.
Loading the software was the usual pain. Most applications I now use are either downloads or sit on the Web, so popping in the DVD seemed almost old fashioned. There followed the usual round of closing running applications, deactivating anti-virus programmes, giving permissions, entering serial numbers, registering this, restarting that, sacrificing a goat, drawing arcane runes, etc. that I almost forgot what the deuce I was installing. Finally, I got it up and running. Luckily, my monitor already has a microphone installed, so I didn't have to use the cumbersome headset that came with the disc. I have no idea where they got it from, but looked an fit like something from the '90s and wasn't a patch on the tiny microphone/earpiece that I use for online meetings when feedback is frowned upon.
Now I had to teach it how to recognise my voice. This was easier than it might have been because the designers preloaded various accents, so all I had to do was click "British". Unfortunately, there was no Yorkshire via Oxbridge and a stint in rep theatre, so there was still a ways to go. I then had to read some text. Again, the designers were thinking ahead and offered a choice of reading. I picked a Dave Barry piece that would have been very entertaining if I hadn't had to re-read half the sentences several times over for the computers benefit.
At last, the software was configured, up, and running. I propped the reference card in front of my keyboard, and off I went.
Theoretically, you should be able to use NaturallySpeaking to control most functions on you computer. You can tell it to open programs, search the Internet or your computer, move your mouse around, and check your e-mail. Yes, you can do this, but it's hardly practical. The software misunderstands at least a quarter of the time, moving the mouse takes several steps, and the entire process is maddeningly slow. If I'd sprained both my wrists and couldn't type, I'd definitely find the function useful, but otherwise it's hardly worth it.
The other major function of NaturallySpeaking is that you can use it for dictation. Turn the microphone on, open any word processing programming such as Word, OpenOffice, Google Docs, or even an address field and you can tell the computer to write down whatever you say. At first, I thought this could be brilliant. I can type faster than I can talk (and a lot more coherently), but the idea of getting on a roll and having the ability to just dump what I'm saying on the page was tempting.
First the good news: NatruallySpeaking actually does a decent job of understanding what you say and writing it out. Speak clearly and naturally and it will write as happily as an stenographer. It will even format certain things such as dates, addresses, currency, and hyphenate some familiar phrases. This is good. However, when it screws up, it really screws up and you end up with sentences that you'd better remember because you'll never figure it out from what the computer wrote. Also, the programme doesn't punctuate or capitalise, so you have to dictate like you're talking to a typesetter or composing a telegram–remembering to pause and say the words properly or it will write out "comma" and "full stop". It also allows you to spell out words, but since it has trouble telling eh difference between letters and words the results tend to be interesting.
You can, theoretically, also use the porgramme to move around the document, cut and paste, format, and so on, but this is a very difficult process that is hardly worth the effort. I found it much easier to do these tasks by hand. This is complicated, however, by NaturallySpeaking's worst problem, which is how slow it runs. I tried rattling off some copy and soon found that the computer was a page behind me. This can be annoying while working, but it's infuriating when you think the computer simply didn't hear properly. Move the cursor at this point and you'll suddenly find NaturallySpeaking dropping text where it was never meant to go.
So, a complete wash, right? Actually, no. Having played with NaturallySpeaking for the better part of a week, I discovered that it does have a practical application. Sometimes I find myself having to include text that it's either impractical or impossible to scan into an OCR application (handwritten notes are an obvious example) and I loathe having to prop my notebook against a paperweight and type all my scribbles in. With NaturallySpeaking, I can now just read them into the document in question and edit them. It's hardly makes it the killer app, but it saves me a tedious chore. I also find it comforting that I have a back up in case a window crashes shut on my fingers.
Bottom line: A fun toy, has some small uses for the experienced writer. I do imagine, however, that for all its limitations it may find its market among the hopeless hunt-and-peck keyboard jockeys.