Slashdot Mirror
Mac Version of NaturallySpeaking Launched
WirePosted writes "MacSpeech, the leading supplier of speech recognition software for the Mac, has canned its long-running iListen product and has launched a Mac version of Dragon NaturallySpeaking, the top-selling Windows speech recognition product. MacSpeech had made a licensing agreement with Dragon's developer, Nuance Communications. The new product is said to reach 99% accuracy after 5 minutes of training."
I'll have to play with Dragon at some point; I just haven't gotten around to it yet. Aside from accuracy errors, the primary issue that bothers me about speech recognition solutions I've tried is the general lack of being able to recognize speech that seems natural to humans but isn't what the system is expecting as input.
This is especially true with over-the-telephone solutions. For example, I am with Rogers Wireless carrier here in Canada, and their automated customer service system prompts you for your phone number. My last 4 digits are 2125, and it is very natural to say "twenty-one, twenty-five" when giving the number to a human being. The speech system, unfortunately, is only sophisticated enough to understand one-digit-at-a-time mode, so you have to suffer through saying "two one two five". Which isn't truly a big deal, but it's frustrating having to learn each system's unique quirks and limits. I suppose the same can be said of any technology.
Oral dictation (as opposed to fixation) is frustrating at best. Punctuation is a critical item that I can't stand dealing with. Trying to get the goddamn software to insert commas and semi-colons can be difficult enough, let alone wanting to actually insert the word "comma" into a paragraph. Then there's trying to spell out acronyms (aka "aka"), or inserting the contents between and including those parentheses. Until dictation of a document can be done with truly minimal correction and post-editing, and can be spoken at a very comfortable pace, I will stick to a keyboard.
Of course, the most entertaining aspect of watching someone else play with speech recognition is the inevitable habit of sounding completely unnatural while speaking. The monotone voice and sounding like a robot are bad enough, let alone those who think that shouting or talking ree... aaa... llll... lllyy... sloowwwww.... llly is going to help. The funniest I've seen was a woman who seemed to think that talking in cutesy baby-talk would win the system over to her side. :)
I just want a system that responds to commands via a programmable keyword. Only when speech recognition is Star Treky enough to respond to its name will I be happy. My computer will be named Minion.
David Weber http://www.baen.com/author_catalog.asp?author=DWeber uses voice recognition software for writing novels.
David talking about it back in 2002.
"On a more technical from I began using voice-activated software when I broke my wrist very badly about two years ago. I've found that it tends to increase the rate at which I can write while I'm actually working, but that it's more fatigue-sensitive than a keyboard. You can push your fingers further than you can push your voice when fatigue begins to blur your pronunciation and confuse the voice recognition feature of your software.
I don't think it's had a major impact on my writing style, but it does affect how I compose sentences. What I mean by that is that because the software prefers complete phrases, in order to let it extrapolate from context when it's trying to decide what word to use for an ambiguous pronunciation, I have to decide how I want a sentence to be shaped before I begin talking to a much greater extent than I had to do before I began typing."
http://sfcrowsnest.co.uk/features/arc/2002/nz5718.php
I was at the Apple Dev conference in 1999 (or so) when the CEO of Dragon got up during Steve's keynote and announced that they were going to develop a Mac version of Dragon.
Almost 10 years later - and it's finally here!
Or at least a follow up announcement is here.
Yes, and to follow along the same line of thought, nobody has ever come out with anything like a speech recogniser designed for programming. Personally, I always figured that a good speech recognition system for both text and commands would need to make use of sounds that don't occur as text. So, you could do something like a special double-whistle to enter command mode, or honk like a goose for undo. Likewise, you could use gibberish words as commands instead of "delete that."
Obviously, it violates the principle that all computers you can talk to should work like Star Trek. But, it seems that just like a command line interface, a spoken interface could be fantastically useful if only somebody would decide that the operator will need some instruction in a few special arcane incantations.
Then, all we'll need is an extension to C so that function prototypes include a way to express the pronunciation of a function name, so a spoken interface IDE could use something like intellisense to parse the API I am using and away we go.
This software's history includes jail terms. Speech recognition has gotten an extremely bad reputation for being worthless garbage, maybe because it is worthless garbage.
Even a 0.5 percent recognition failure rate is enough to make speech recognition software worse than worthless. The reason is that speech recognition software never makes a spelling mistake. Instead, the mistakes are often extremely difficult to recognize, and sometimes change the meaning in subtle ways. That's partly because when the software is confused it tries to select something that is grammatically plausible.
The result is that it has become difficult to sell speech recognition software. A high enough percentage of people in the U.S. culture know that it isn't actually useful. The orginal owners of Dragon NaturallySpeaking sold the product to a company that sold it to the company that became Nuance, maybe because they felt the product was damaging the credibility of their trademarks.
Here is a quote from the ComputerWorld story linked in the earlier Slashdot story, Is Speech Recognition Finally 'Good Enough'?:
"In 1993 two executives from Kurzweill Applied Intelligence (which pioneered SR for the medical market) went to prison for faking sales. That firm was sold in 1997 to a Belgium SR firm, Lernout and Hauspie (L&H), which was reporting phenomenal sales growth at the time. Dragon Systems, which originated DNS that year, was reporting only anemic growth, and L&H had no trouble acquiring Dragon Systems in early 2000 in a stock deal. Within a year a series of accounting frauds came to light and L&H collapsed into bankruptcy. Its SR technology was sold in late 2001 to ScanSoft Inc., which kept the DNS line going. (It was then at Version 6.0.) ScanSoft later acquired Nuance and adopted its name.
"Thereafter, "It was with the launch of Version 8.0 (in November 2004) that the market became reinvigorated and took off," said Chris Strammiello, director of product management at Nuance. "We crossed an invisible line with Version 8.0, where the software actually delivered on its promises and offered real utility for the users. Sales have been growing at a rate of 30% yearly since then, except that we expect it to do better than 30% this year."
Read that again: "... the software actually delivered on its promises and offered real utility..." I called Nuance and was told that version 8 did not have a new recognition engine, but only had improvements in the user interface. A friend who owns and tested version 8 told me he could see no difference in accuracy between that and version 7.
So, in my opinion, Nuance has done common deceitful things that are called "Marketing":
1) Bring out new versions. Previously, when there has been a "new version" of Dragon NaturallySpeaking, I call Nuance technical support and ask if there is a new recognition engine. I didn't call for version 9, but for the last two versions they have said no. So, nothing is changed; the software is still worse than useless to me, in spite of the fact that they advertise that the software is now more accurate.
How is it possible that the software is more accurate, if the recognition engine did not change? Maybe it isn't true. Or maybe the company improved the guesses the software makes when the software really has no clue what the user said. As I mentioned, those guesses have become so sophisticated that you can become confused about what you actually said, and you have to spend time re-creating your ideas. If you are saying simple things about a simple subject, this is not as much of problem as when you are writing about contract negotiations, for example.
In the words of a Slashdot reader: "The opinions expressed here may be those of my speech recognition so
I am not a native English speaker, but I am usually able to say just about anything I want. In this comments, I have not altered any of the mistakes (if any) that Dragon NaturallySpeaking made while I was dictating. As you can see, the error rate is probably a bit higher than 99 per cent correctness. Nevertheless, I used this extensively, because it increases the speed at which I can work.I often have to type reports, and it goes a lot faster while using this tool. The only problem is that these reports contain lots of enterprise specific (and IT specific) terms. Naturally, it takes a while before Dragon NaturallySpeaking knows all of these terms.
Other than that, I am very happy with it.
I have a photographic memory for numbers. I know almost a hundred of them.
It's fine to port this to the Mac. Fine. Good. Whoopie.
But they are so DROPPING THE BALL. They have the best voice-rec platform. (You can think it's not good enough, but it's still the best.) What they need is to port it to Linux. Duh! Wake UP!
No, I'm not just saying the usual "Does it run on Linux?" bit. Linux is the now (and coming even more) obvious OS for small devices. When you want to talk to ANY device in your home or car, or your cell phone or PDA, you'll be talking to LINUX. THAT'S where we need a great voice-rec system. We need it ported to Linux and opened for an API. This will catapult this annoying desktop app into a present on almost everything type software device in a matter of a couple of years -- as low power devices provide enough umph to do what the heavy machines of a few years ago do.
I'm a radiologist who uses a Nuance product for several hours a day, every day, and my experience has been overwhelmingly positive. Whereas I used to waste a great deal of time editing and correcting mistakes by human transcriptionists, I only occasionally have to manually correct the Nuance transcriptions. Our throughput and efficiency have increased considerably since we started with the product, and there is absolutely no way that I'd ever return to the previous system. The adoption of speech recognition has been the biggest advance in my field since digital imaging, IMO. Oh, and "when the software is confused it tries to select something that is grammatically plausible"? I don't think so - the software has no concept of grammar.