Why some have no choice but to use ViaVoice
Submitted by floridagal on Mon, 08/27/2007 - 15:58.
There's a thread that's been locked, so I can't reply. Someone had questions regarding ViaVoice, and as usual, someone else posted the "Why use that outdated software - upgrade to Dragon" answer.
Let me enlighten you. Some people are using it to provide captioning. You've seen captions on TV. ViaVoice is the only SRE that gets the words to show up quickly enough for that. You can't have blocks of words appear all at once. Television stations and captioning companies have standards whereby the words have to appear on the screen within X number of seconds of being spoken. Dragon cannot meet those standards.
ViaVoice is also being used to provide CART services: Communication Access Realtime Translation. That could be anything from sitting in a classroom with a hearing-impaired student who sees the words of the professor and classmates appear on the provider's laptop, to providing the realtime text projected on a large screen in a church or seminar. It's all about access for the deaf and hard of hearing. Most of that is being done by stenotype court reporters turned captioners and CART providers. But there aren't enough stenocaptioners to fill the need because it takes on average 4 years, a boatload of money for the schooling, and the dropout rate is about 80%-90%, and reaching that level of speed and accuracy necessary for CART and captioning is extremely hard.
So, while ViaVoice might be a dead end, it's the only product available right now for "voice writers" - those using speech recognition to provide captioning and CART. The words need to stream, not drop down in blocks, in order for the recipients to read and comprehend in a live environment.
If you decide to turn the captions on your TV to see what I'm talking about, choose something like CNN which has quality captions. Don't choose the Home Shopping Network, HSN - they are using voice writers who have had inadequate training and are currently the embarrassment of the captioning industry.
I just commented (again) on
I just commented (again) on that thread, I'll summarise my comment here. NaturallySpeaking works with any length of utterance you choose. It's true that the text will 'drop down' as opposed to 'stream' but that's a technicality given 'any length of utterance'.
I add that -- see my earlier comments -- I really don't want to take anyone away from VV, I am simply trying to help out.
Judy
Ah, and I just responded to
Ah, and I just responded to you on that thread! It would make everyone's lives easier ("everyone" being current and future voice writers) if Dragon could be used as eventually no one will be able to purchase a computer with XP and therefore ViaVoice won't be an option. This week I'll be setting up my CAT software with both Dragon and ViaVoice so I can do my own analysis. But right now, the standard for voice writers is ViaVoice for captioning and CART, and Dragon for court reporting, because in that environment, the recipients of the realtime feed - judges and attorneys - don't read the text as it's spoken, but rather use the text for review to formulate questions or rule on objections.
Hello Floridagal and I just
Hello Floridagal
and I just replied to you
It would make everyone's lives easier ("everyone" being current and future voice writers) if Dragon could be used
I agree entirely
This week I'll be setting up my CAT software with both Dragon and ViaVoice so I can do my own analysis.
oh great, thank you! I've been trying to replicate what you do but of course I don't have CAT software so can't, really.
I'd noticed a lot of court reporters used NaturallySpeaking, I didn't know why (I'm used to people using it, of course)
Judy
Judy Evans wrote: I just
I just commented (again) on that thread, NaturallySpeaking works with any length of utterance you choose. It's true that the text will 'drop down' as opposed to 'stream' but that's a technicality given 'any length of utterance'.
I just conducted a simple experiment. I took a deep breath and, at a steady pace, I dictated a complete sentence without pausing. I did this first with Dragon NaturallySpeaking preferred version 9.5, and then with ViaVoice version 10.5. The sentence was as follows:
"Four score and seven years ago our fathers brought forth on this continent a new nation conceived in liberty and dedicated to the principle that all men are created equal."
With DNS, nothing appeared on the screen until I had completed the sentence, at which point, the entire sentence appeared almost instantaneously. With ViaVoice, words started appearing after I had uttered the word "ago", and the rest of the sentence appeared in three or four words segments. My speech was never more than three or four words ahead of the screen.
I then repeated the experiment changing only my style of dictation. Instead of continuous speech, I spoke discreet words at a steady pace but with a very slight pause between each word. This time both programs yielded similar results. The words on the screen appeared in phrases of from two to four words in length. I think the phrase is produced by ViaVoice may have averaged fewer words than those produced by DNS but I'm not sure. I also think that ViaVoice took slightly less time to get all the words on the screen but I'm not sure about that either.
I will not draw conclusions or attempt to extrapolate. I do not know how many people could repeat my results, nor do I know what the significance would be relative to voice captioning if the results were repeatable.
Charles
Charles Devonshire wrote: I
I spoke discreet words at a steady pace but with a very slight pause between each word.
Correction: I discreetly spoke discrete words at a steady pace but with a very slight pause between each word.
Charles
Charles I haven't got Via
Charles
I haven't got Via Voice.
As you'll see from my posts in the other thread
http://www.speechcomputing.com/node/1498#comment-6273
I can dictate a sentence in about 8 bits, 2-3 words each; that's actually less hassle, puts less strain on the voice, I'd say, than 'discrete dictation'.
The words on the screen appeared in phrases of from two to four words in length.
I get that if I dictate in short phrases; if I dictate discretely i.e. pausing after each word, the words appear one word at a time. (I did that just now to see what happened.) So, your results puzzle me slightly. That is, the second result does. (The first -- that when you spoke a sentence as an utterance, NaturallySpeaking didn't begin typing it till you'd finished the utterance -- obviously doesn't surprise me, that is the way it works... .)
But I have no doubt that NaturallySpeaking can type a sentence in bursts of 2-3 words (or in phrases...), because I know it can. I normally dictate that way -- not 2-3 words, but in phrases -- and get very good results.
It seems, though (see Floridagal's posts here and in the other thread) that captioners prefer ViaVoice a) because it's linked with the captioning software b) because it types onto screen immediately, word by word.
Judy
Charles Devonshire wrote: I
I just conducted a simple experiment. I took a deep breath and, at a steady pace, I dictated a complete sentence without pausing. I did this first with Dragon NaturallySpeaking preferred version 9.5, and then with ViaVoice version 10.5. The sentence was as follows:
"Four score and seven years ago our fathers brought forth on this continent a new nation conceived in liberty and dedicated to the principle that all men are created equal."
With DNS, nothing appeared on the screen until I had completed the sentence, at which point, the entire sentence appeared almost instantaneously. With ViaVoice, words started appearing after I had uttered the word "ago", and the rest of the sentence appeared in three or four words segments. My speech was never more than three or four words ahead of the screen.
Charles
Charles,
As I noted, I think this behavior occurs because DNS 9 uses more context (i.e., the surrounding text) in the interests of greater accuracy. I'm confident what floridagal says about the behavior of VV is accurate -- its an older program without the up-to-date technology embedded in the latest version of DNS.
Its a trade-off between speed and accuracy for the sake of accuracy for primarily text writers rather than realtime writers.
We have to hope that DNS' developers could add a control or two so that realtime writers can achieve their preferred behavior. As noted by Knowbrainer, adjusting the speed-accuracy slider will go a way towards this goal. Speaking in shorter utterances, as you did, will also help, but I'm not sure that is compatible with the best interests of the realtime writers as I understand them -- they want maximum dictation speed without worrying about chopping up their dictation into two- or three-word groups!
Another alternative would be to produce a special version for the realtime writing market, perhaps sold only to vendors who use it as a front-end to their back-end processors.
I may be wrong, but it seems to me as floridagal has commented, that DNS is flexible enough for court reporting usage.
Bruce
floridagal, Please note that
floridagal,
Please note that we are open to the discussion of problems with any SR product! We understand there is anxiety among long-term ViaVoice users, but the consensus even among us non-SR retailers is that its bad faith to PROMOTE a product to newbies that has no future UNLESS you carefully hedge that promotion with information about the product's viability.
That's the problem we have with a particular VV partisan, who for all his good work dogmatically exhibits bad faith by suggesting new SR users try a product whose development has ceased and will not work with an expanding range of softwares.
Its pretty simple: No bad faith promotion, no controversy. In the SR community we feel newbies need some protection in the form of full-information disclosures. Then they can make an informed decision about what software to choose.
For current VV users, this forum welcomes discussions of how to cope with any problems that might arise.
Bruce
For current VV users, this
For current VV users, this forum welcomes discussions of how to cope with any problems that might arise.
yes indeed. It might help if more of us took part (I'm thinking of myself, Bruce!)
Judy
Bruce Cyr wrote: That's the
That's the problem we have with a particular VV partisan, who for all his good work dogmatically exhibits bad faith by suggesting new SR users try a product whose development has ceased and will not work with an expanding range of softwares.
As there is no doubt that I am the " particular VV partisan" to whom Bruce refers, I feel I must reply.
There have been arguments on this site as to whether or not ViaVoice is a dead programme, or still viable. I have maintained, and still maintain, that it is viable in all OS systems up to XP. While it is no longer apparently being developed, it is still, as seen from this thread, the favoured SR programme for certain contributors' type of use. The only thing which appears to affect its viability is the coming into operation of Vista. As Englishem on the sister thread has stated, she got 99 per cent of ViaVoice to install in Vista. I am certain that one per cent can be overcome as soon as it is established what is causing the block.
Bruce, I object your reference to me exhibiting bad faith. At no time have I ever encouraged anyone to do what they did not want to do. I have not yet come across any programmes in which ViaVoice will not work. However that is only my experience. I'm sure there may be some where this may hold true, but for you to make this statement without having the benefit of using ViaVoice is a nonsense.
In the meantime, for anyone using XP and Microsoft Office up to 2003 version, ViaVoice is a viable (and inexpensive) option for SR and will continue to be so.
Quentin
I concur with Bruce on this
I concur with Bruce on this issue. I find myself frequently posting after you because many of your posting make it sound like ViaVoice is the way to go and I understand it's your favorite speech recognition program but I think your enthusiasm can be misleading because you rarely if ever mention, or even admit, that ViaVoice will soon be discontinued, does not support Windows Vista or Office 2007. Potential customers need to hear the pros and cons. We don't preach that NaturallySpeaking is a godsend. If you visit the KnowBrainer Speech Recognition Forum you'll find we and others have a number of “issues” with NaturallySpeaking. You have accused me of pushing NaturallySpeaking for-profit reasons but I'm also licensed to sell ViaVoice. I don't because I feel it would be an injustice to the consumer. The only time I make any real profit selling speech recognition software is what I sell a full boxed copy of DNS Pro 9. I make $14 on the Preferred 9 upgrade but include 30 days of unlimited virtual on-site technical support which means I potentially lose money on every sale. I sell the full boxed copies of Preferred 9 for $1 above my cost. Because of my licensing agreement, I am required to purchase from distribution and don't get the great discounts that non-licensed vendors get from Nuance. I still include technical support and a training guide at no additional cost and because I take credit cards, I lose money on every sale. However I do make a profit when those customers return to purchase KnowBrainer 2006, microphones and digital recorders.
Lunis Orcutt - Developer of KnowBrainer (DNS Command Software)
A Nuance Gold Certified Endorsed Vendor
crivon1 wrote: As there is
As there is no doubt that I am the " particular VV partisan" to whom Bruce refers, I feel I must reply.
Quentin
Quentin,
As the confessed partisan, you might want to clarify exactly why it is that you propound the use of VV without providing full disclosure of its condition.
For example, is it your goal to convince enough people to switch to VV to convince IBM to revive it?
That's what baffles me -- because it would be well-nigh impossible for IBM to revive it as a desktop product given what they have sold or licensed to Nuance and the current competitive environment.
In any case, if you were to recommend it while admitting it is a dead-end product with no hope of future development, that would probably quieten the "controversies". In fact, you could use VV's text macro capabilities to generate a boilerplate paragraph or two that does both things in a full-sweep, thereby saving yourself the burden of redictating the text every time the opportunity presents itself!
Bruce
We might be able to shed a
We might be able to shed a little more light on the ViaVoice semi-acquisition by Nuance.
We don't believe that IBM was allowed to outright sell to Nuance because of US monopoly laws but they were able to skirt the issue by selling distribution rights and all profits which would effectively prevent IBM developers from improving or even caring about ViaVoice because they couldn't profit from it. Nuance probably thought this was just as good as purchasing full ownership and copyrights of ViaVoice up until the point where IBM donated the ViaVoice code by making freeware. It is our understanding that the code is freely available to anyone who wants to develop it but we haven't heard of any third parties doing so.
In order for any third-party developer to voice enable a product with NaturallySpeaking they are required to purchase a DNS Pro license. Nuance will not permit third-party vendors to use their SDK with DNS Standard or Preferred which should have increased the incentive for third-party developers to jump on the ViaVoice code but it's also possible that everyone was worried about Microsoft (the 800 pound gorilla) announcing that they were going to include a speech engine with their soon-to-be released Vista OS.
Lunis Orcutt - Developer of KnowBrainer (DNS Command Software)
A Nuance Gold Certified Endorsed Vendor
BruceCyr wrote: As the
As the confessed partisan, you might want to clarify exactly why it is that you propound the use of VV without providing full disclosure of its condition.
This "condition" in what you and others on this site maintain. In the meantime Nuance, and retailers in the UK and Ireland (I am not aware of what is happening in the rest of Europe, Asia, Africa or Australia) are currently marketing and selling ViaVoice. If a product is dead, why market and manufacture it?
For example, is it your goal to convince enough people to switch to VV to convince IBM to revive it?
Why not? IBM is still using ViaVoice technology in its various voice products. Since they appear to hold the source codes, what is to stop them developing?
-- because it would be well-nigh impossible for IBM to revive it as a desktop product ......
Why not?
Just because the development of a programme has ceased does not make it a programme that cannot be used successfully. For example what is wrong with Office 97, or Windows 98? There are very many satisfied users of these so-called dead programmes. Personally I and my co-professionals in my office, are very happy with Office 97, and this is also used by our secretarial staff very successfully with Windows 98.
Whilst many of the subscribers to this site are technophobes, I would venture to opine that the majority of readers would not be that interested in, or even understand, the high level technical debate that goes on here. They would be more interested in finding solutions to basic problems, or getting ideas of what more they could do with the programmes they use.
Quentin
crivon1 wrote: Whilst many
Whilst many of the subscribers to this site are technophobes, I would venture to opine that the majority of readers would not be that interested in, or even understand, the high level technical debate that goes on here. They would be more interested in finding solutions to basic problems, or getting ideas of what more they could do with the programmes they use.
Quentin
You meant "technophiles"?
With that lawyerly display of sophistry you've highlighted why your heedless advocacy will continue to evoke controversy: The SR community has to protect the best interests of its newcomers by insuring that they understand all the conditions and caveats.
I'm not sure everyone appreciates the grandeur of your spectacle as I do -- you are the specimen Don Quixote-like reliquary in our too prosaic neck of the woods! Its like finding a stranger's hair in your soap bar!
Bruce
Of course I meant
Of course I meant technophiles. It was an early morning slip of the tongue that made me say technophobes, and I never noticed it on proof-reading.
However it does not take away from the general tenor of what you quoted.
Quentin
Bruce, I completely
Bruce,
I completely understand what you're saying. But voice writers are in a different category as compared to the people you probably usually see on this forum. I don't know if you read my post in the other thread, but ViaVoice and Dragon aren't being used alone. They're being used in conjunction with CAT software, and the CAT software companies have configured their software to utilize the input from ViaVoice and Dragon. The only SR software you can use is that which the CAT software has been designed to interact with.
The people who advise ViaVoice for captioning and CART are the pioneers in this application of speech recognition, those who have used both Dragon and ViaVoice and are the few providing quality voice realtime. And everyone knows that ViaVoice hasn't been updated and it's unlikely it ever will be. I suspect that the CAT companies will be instrumental in seeing to it that another option is available once ViaVoice is no longer useable. But once you develop your skills at dictating correctly for 96% or above accuracy at high rates of speed (180-200 wpm at 96% is the minimum standard for realtime), you can apply that in the creation of a new voice model with whatever software comes down the pike. Of course, there's a lot more that goes into it, such as inputting brief forms - as an example, saying "smaft" for "as a matter of fact" - and preparing and running the documents through the software for context. For instance, I say "kah" when I want the punctuation of a comma. Every document has to be prepared so that all commas are replaced with "kah."
Everything you ever wanted to know about voice writing but were afraid to ask?
floridagal
Everything you ever wanted to know about voice writing but were afraid to ask? :)
I find this interesting. Sometimes I turn off the audio on the television and just read the captions. It's very handy when I'm on a telephone call and want to see what's happening on a news show or similar.
There are some channels that have extremely poor captioning some are pure rubbish (usually symbols and random letters). I've seen a few quality feeds and enjoy those.
I'd stay with whatever worked the fastest! I wonder if a custom or perhaps smaller vocabulary from Nuance could help speed up DNS? Do you know if anyone has used the Microsoft Vista Speech Recognition system for this?
Skip
P.S. I also created a forum just for captioning and added categories for professions. A post can be placed into multiple categories under professions, just hold the control key while you select them. I've not tried this by voice yet, hopefully it will work. Someone let me know when you try it!
I appreciate your adding the
I appreciate your adding the captioning forum! I'll have to let others know about it.
If you see symbols and random letters in captions, that's a technical issue having to do with the decoders and whatnot. I'm don't know enough about that end of the technology to be more specific.
If you see wrong words, that's the captioner's fault. For stenocaptioners, one misplaced finger can create a different word than is intended. I can read through it, though.
One of the CAT companies is planning to make their software compatible with Microsoft's speech recognition. I don't know about the others.
As a matter of fact, Dragon without the bells and whistles would work better for us. My CAT software works with either Dragon Pro or Preferred. I use Preferred because I don't need multiple vocabularies. You can accomplish that task in Preferred by creating a base user, backing it up, and restoring it with a different name and making it situation- or topic-specific. That's how it's done with ViaVoice. The reason you'd want to do that is, for instance, you do both court reporting and CART. Court reporting, with its Q and A format, is very different than, say, a college class or church service.
Using Dragon and CAT, you don't need commands mode. CAT takes care of all the formatting, so it's always used in dictation mode. In fact, if you forget to make sure it's in dictation mode, things get totally screwed up.
Also, the first thing I have to do when creating a new Dragon user is delete all the punctuation (because I use voice briefs for that so as not to conflict with the word, and it's faster to say "kah" when you want a comma, and the CAT software is programmed to know kah=,) as well as things that are not used in court reporting, captioning or CART, such as OK for okay, most abbreviations, all of the commands. It would be so much easier to just add what I want than to take out everything I don't want.