Wednesday, March 26, 2008

Dictation software

I've tried a few times to start using dictation software. If I recall correctly, the first time was right after the millennium, with the first edition of MacSpeech iListen software for the Mac. It didn't work too well. A few years later, I tried buying Dragon NaturallySpeaking for Windows for my old Windows laptop. This did not work very well either. I don't know whether the Windows laptop was underpowered, perhaps. Last year after getting an Intel Mac I bought Windows Vista for it, and a new version of NaturallySpeaking. Unfortunately I could never get that the dictation software to install correctly. David Pogue told me that his worked well but he was using Windows XP. I did not have the energy to purchase XP also.

But then MacSpeech finally managed to purchase a license for the speaking engine of NaturallySpeaking. And so we should finally have decent dictation software for the Mac, which has been pretty much the only thing I have been missing.

I got the software yesterday. And this whole post, as well as half of the last one, has all been dictated in MacSpeech Dictate for the Mac.

I'm still learning how to use it, and maybe the software is too, so I still have to make quite a few corrections. But it is usable already, and it's fun, and I'm looking forward to seeing how it will go in the future.

Update: well... so far, not so good. There are corrections in at least every other sentence. Boring. I don't get how other reviewers are getting the 99.5% correct dictation that they speak of. (And I've done all three voice-training texts provided.)

9 comments:

Anonymous said...

"I got the software yesterday. And this whole post, as well as half of the last one, has all been dictated in MacSpeech Dictate for the Mac."

Interesting. Perhaps this could be used to dictate a book. You could then pay someone to go over the broken English, and maybe have Lulu.com or someone distribute the product to the masses.

Instant Stephen King.

Eolake Stobblehouse said...

Yes, this has been one of the reasons I've wanted it.

I'm a little disappointed, though, the software does not appear to have any button named "write novel".

Anonymous said...

Of course not. MacSpeech, Inc. would be shooting themselves in the foot if they had that. For what would you then need the speech recognition for?

You need to say a novel.

Bert said...

When you are having a real-life conversation, you have immediate and direct feedback from your audience, and are free to use as many words as needed to convey your ideas. Sometimes, when facing obvious incomprehension, you will change your angle completely in an attempt to reach the other.

Such freedom is completely absent from the written word, which must convey your ideas with maximum efficiency on the very first, and usually only attempt. Because of this, effective written communication requires a lot more thought and editing work to be efficient.

For my part, I can only perceive dictation software as yet another barrier between me and the reader, and one that promotes laziness on top of that.

Sure, it may be a fun and efficient toy in an informal context like this blog, but to write a novel, or any form of book for that matter?

Not for me. You haven't read a single sentence in this post that wasn't changed in some way or another.

Eolake Stobblehouse said...

Really? Interesting. I'm pretty much a first-draft writer.

Alex said...

The one that surprises me is that in all the flavours of Naturally Speaking (Medical, legal etc) there is not a "Programmer" version.

I guess coding needs voice macros more than a natural language Speech to text tool.

As for writing needing to be bang on first time. This is not the case in the whole world. I have had some docs go through 15 or 20 review cycles before they are acceptable.

(Word verification "holdm"; hold 'em?)

Bert said...

"I guess coding needs voice macros more than a natural language Speech to text tool."

My guess is that they never could match the speed of typing keywords and symbols (compare saying "open parenthesis" with typing "("). Not many languages where dictation would work. I'd love to see an APL version of that, too. LOL!

[APL has to be the most cryptic language ever invented. A typical statement looks like this: "X[⍋X+.≠' ';]" (without quotation marks). Note that this statement is exceptional, as it can be reproduced here. To write APL, you need a special keyboard, for most of the symbols don't exist anywhere else. Anyway, sorry for the geeky humor.]

"As for writing needing to be bang on first time. This is not the case in the whole world. I have had some docs go through 15 or 20 review cycles before they are acceptable."

Makes my point obvious, in a way. As a freelancer, I never could make a dime if my writings had to go through more than one, maybe two reviews. Twenty reviews has to be either military, gov. or some other "no useful work expected here" outfit.

Alex said...

20 revs for specification of freeway traffic count collection via cellular motorist aid callbox, with additional data filtering for incident detection.

Metropolitan Transportation Commission.

15 rev, specification for multiple DSPs and RISC engines in SOC ASIC.

In these instances everything has to be explicit, self contained and unambiguous.

With an RSI even saying "function declare foo, byte x, word y end body" would be less painful than trying to span the keyboard to get the braces and parenthesis needed.

Bert said...

That's what I said, MTC is gov.

XXX bus protocol and API specification.

Links hundreds of devices from dozens of vendors. Worldwide diffusion. Two revisions. ;-)

Many others, but all under NDA, no can tell.