The great dictator

Keith White
18 December, 2008
View more articles fromthe author
AAA
Blogs

Some time ago I promised our readers to take MacSpeech Dictate for a test run, but I’ve been waiting for version 1.2 because it has three major new features worth waiting for. Move/Insert commands allow me to zip around in a document without touching the keyboard; a new Spelling Mode makes it easier to spell out names and acronyms; and Phrase Training, similar to Correction Mode in other voice recognition programs, helps me quickly hone my recognition accuracy.

Firstly I must confess I am not completely new to all this. As a keen user of IBM’s ViaVoice software for many years I found installation and then initial training, by reading short stories into a correctly set-up microphone was a breeze. The main difference was how accurately Dictate was able to recognise my voice after only five minutes’ training. In ViaVoice this process had taken a good half hour, but then, that was technological light years ago.

From previous experience I remembered the need to speak clearly but not overly, and to try to speak in phrases and sentences rather than discrete words. Out of the box Dictate somehow seemed to know the names of a lot of people at the school I work at plus a random selection of Australian politicians I threw at it. Try – Jeff Kennett Kim Beazley Kevin Rudd Paul Keating John Howard and Malcolm Turnbull. It got that last sentence first go! And I hadn’t taught it any of those names. Not to my knowledge, anyway.

Dictate has two major functions: typing as you speak into it, and controlling your Mac by voice commands. At present, I’m not particularly interested in voice control –  I can’t stand the feel of headsets — but I did give Command Mode a quick workout. It all seemed to work pretty smoothly, opening and closing applications and documents and pressing buttons in dialogue boxes.

Being a total keyboard klutz — I can make three mistakes typing one word — I am more interested in any means by which I can avoid typing. And happy to invest time in achieving this.

In ViaVoice I was able to move between voice and keyboard input at will. This is not recommended in Dictate, as it interferes with the software’s recognition of a document’s structure. Particularly voice navigation through the document. As they say, if you’re talking, talk; if you’re typing, type. I was already familiar with creating new lines and paragraphs, but now I can go to any word in my document with Insert and Move. It took me a bit of retraining to keep my hand off the mouse but after a solid practice session I felt comfortable with all the moves.

Dealing with misrecognised words generally begins with a basic command “scratch that” or “forget that”. This simply gets rid of the last thing you said and quickly allows you to have another go. I found this worked pretty well for me. Otherwise I have a transparent Recognition Window which shows me a number of possible variations from my growing voice profile of what I just dictated. If I find the correct version among these I simply tell Dictate to choose the phrase number ‘x’. If I don’t see a phrase matching what I wanted to say then I can edit the nearest version in the Recognition Window and then get Dictate to insert that corrected version into my text. This correction is then stored in my profile for next time. I’m quite happy to do this as I know it’s going to save me time in the long run.

I can enter Spelling Mode at any time by saying just that and then return to Dictation or Command Mode similarly. Dictate can handle quite a large list of punctuation symbols from & to ~. I tested them all without a hitch. Numbers were pretty well handled too, except for a problem with some fractions which I solved by using the alternative “slash” instead of “over”. Roman numerals were a breeze – X out of X. Date and Time needed a little fine tuning but by the end of my first session Dictate was doing what it was told.

So, after one serious day with Dictate 1.2, I’m pretty happy. I’ve moved all over this piece without a mouse, used the Recognition Window to correct misrecognised words, and got capitalisation and numbers in the bag. I’ve also trialled another useful feature called Vocabulary Training. Here I simply drag text selections from documents I have written into the Dictate WordPad window and then tell Dictate to train the vocabulary from my selection. Dictate then shows a list of words it doesn’t already know and gives me the option of adding them to my vocabulary. Simple.

MacSpeech Dictate 1.2 has already impressed me with its speed, stability and accuracy and there are a raft of other features that haven’t tried yet. I think we’re going to be friends for a long time.

For this exercise I stuck to WordPad, Dictate’s native document format as this is recommended to provide the most stable and accurate dictation environment. But I’ve noted already that it works OK in Pages and Text Edit. More about this and other features next month.

Leave a Comment

Please keep your comments friendly on the topic.

Contact us