 |
 | Home | Contact Us | FAQ | Search & Site Map | Link to Us |
 | Sign In | Join | Other 45 Sites in Network |
Mac Forum / Applications / Mac Applications / February 2005Voice Recognition, iListen stuck at starting gate
Just for kicks, thought I would evaluate MacSpeech's latest offering of "iListen" version 1.6.4 for OS X. Supposedly, correction of dictation mistakes can be done by voice, completely hands free, no keyboard needed at all. I bought the $150 product that ships with the included "Telex" headset microphone. Plenty of instructions on how to hook up and configure the headset for every flavor of Wintel, but no instructions for a Mac. Booted up the CD, and found the instructions there for Mac hookup. Finally got audio response from the headset mic', but the iListen app' kept insisting that the audio quality was too low for speech recognition. Tried hooking up the microphone directly via the jacks on the rear of the Pismo powerbook, and also tried hooking it up via the USB input, using the supplied USB converter, but same bad result. Tried the advanced manual settings for volume, but same bad result. So far, the only speech recognition app' that seems to be worth a tinker's dam is the high cost "Naturally Speaking Pro" sold by Lernout & Hauspie corporation. As I recall, that costs about $600, and only works with a PC, not a Mac. (medical and legal versions cost $1,000) I used NP-Pro, and hated the fact that I had to use a peecee just in order to get good performance of speech recognition along with the ability to correct mistakes by voice only, no keyboard required. The 'average' performance is about 98% accuracy, with occassional runs of several paragraphs at 100% accuracy. Tried every Macintosh speech recognition app' I could get my hands on, but none of them are worth a crap compared to "Naturally Speaking". Here, I am running a Pismo powerbook with OS 10.3.5 Has anyone else managed to get iListen to work on a Mac? Hope springs eternal. Perhaps some day in the distant future someone will create a quality speech recognition app' for the Mac, one that can compete with the PC software for dictation usage. Even with the PC, I have found that 200 WPM is about the fastest that can be handled with anything near 'normal' 98% accuracy. Voice correction of the remaining 2% of mistakes drops the overall speed to effectively about 40 WPM. Most people's normal speaking rate is around 300 WPM. Mark- Mark Conrade: > So far, the only speech recognition app' that seems to be worth a > tinker's dam is the high cost "Naturally Speaking Pro" sold by Lernout [quoted text clipped - 6 lines] > order to get good performance of speech recognition along with the > ability to correct mistakes by voice only, no keyboard required. I have six or more Macs (some loaned out, never to be seen again) and one Sony Vaio laptop that I keep to run certain astronomy software. I never use the Vaio when a Mac will do the job. So it pains me to say that, in truth, if you want to do voice recognition in a serious way, you're better off using Windows. Even the moderately priced IBM VIaVoice works better on Windows than any speech recognition software for the Mac. The good news is that speech recognition on the Mac is likely to improve dramatically in the next year or so -- but some developer is going to have to step in with a top-end product to take advantage of that. I expect that there would be lots of Mac users behind me in line when such a product is released. Davoud  Signatureusenet *at* davidillig dawt com > So it pains me to say > that, in truth, if you want to do voice recognition in a serious way, > you're better off using Windows. Even the moderately priced IBM > VIaVoice works better on Windows than any speech recognition software > for the Mac. Yeah, I get that impression also. I have an old Compaq PC lying around here somewhere. Think I will try the same iListen software on it, to try to resolve whether the problem lies with Mac hardware. > The good news is that speech recognition on the Mac is likely to > improve dramatically in the next year or so -- but some developer is > going to have to step in with a top-end product to take advantage of > that. I expect that there would be lots of Mac users behind me in line > when such a product is released. I will be one of the guys standing in that line. <g> Mark- > Just for kicks, thought I would evaluate MacSpeech's latest offering of > "iListen" version 1.6.4 for OS X. [quoted text clipped - 17 lines] > > Tried the advanced manual settings for volume, but same bad result. Interesting. I recently started using iListen as well on my Quicksilver which has no built-in mic. I already had a headset so mine was only $99. I get the same message as you though I am using an Andrea ANC 750 headset plugged into a Griffin iMic (USB). But despite the message that the audio quality is barely acceptable and that I continue to speak at the same soft level, I have found that iListen does an awesome job in translation anyway. And it doesn't require a speak-pad, either. Like you, I was rather upset initially over the "poor quality" message but I gave it a chance anyway. I've been using it in MT Newswatcher, Mellel, iChat, Mail, Text Edit and other applications quite successfully--- though I prefer to use a combination of voice and hands for correction since I find that to be faster than either separately. I think it could stand speed and design improvements in the correction mechanism and, every rare now and then it just up and quits on me but I don't know why. I do find correction a bit clunky and slow. However, for me, that far outweighs the awful inconvenience of having to use a speak-pad and not being able to use it in all programs. But what I most like about it is being able to speak at a normal rate of conversation. iListen does better for me when I pause less and talk faster. It does take some time to learn how to use it and to understand how it works. And, I nly use it in OS X. I id not do an OS 9 installation at all. YMMV Madeleine > Like you, I was rather upset initially over the "poor quality" > message but I gave it a chance anyway. I would be happy with "poor quality", however mine was "unacceptable" so it would not allow me to continue the setup routine. What really frosts my pumpkin is a sentence in the user's manual that sez in effect "Now that you have bought our approved microphone headset, you will not have to worry about the quality of the microphone". Yeah, right. <g> Those were not their exact words, but it comes close. I never had these problems when I was using "Naturally Speaking" on a Compaq Presario laptop running the XP operating system, the setup went flawlessly. Think I will try a different headset. I have a "Parrott VX1". Do not know which voice recog' product it came with, I have tried so many of those products. I think that headset came with ViaVoice, another product that did not work well on my particular Mac. Only voice recognition software I will recommend without hesitation is Naturally Speaking Pro, but that puppy costs $600 and is PC only. Among other good features of NS-Pro, I can dictate into a tiny shirt-pocket sized Sony digital recorder model ICD-MS1 when I am away from my computer, then later run the recording into the computer, where recognition accuracy averaged about 98%. (at about 200 WPM speaking rate) In fact, I noticed no difference in recognition accuracy when I used the tiny recorder with its built-in microphone, as contrasted to using the computer and headset combination. About 98% average, with short runs of a few paragraphs at 100% recognition accuracy. What consumes all the time is the correction process. With NS-Pro, there is almost no chance that the resulting "corrected" text will be less than 100% accurate, because NS-Pro keeps your original voice recording for comparison purposes. That means you can compare what you said word-for-word, and correct the written text so that it perfectly matches the original voice recording. As regards the practical uses of voice recognition software, there are very few reasons for using it at the present time. The average recognition accuracy will have to improve dramatically before it will be useful, IMO. Somewhere up around 99.8% in my opinion, or ten times more accurate than they are at the present time. For that to happen, these programs will need to incorporate sophisticated artificial intelligence techniques, which they don't at the present time. What determines the "correct" spelling of a word often depends on the surrounding words, phrases, sentences, and sometimes on the "backgrounds" of the people doing the speaking and reading. (for example, if you were an office worker and had to "correct" a doctor's dictation, you might not know the "correct" spelling of a medical word) Real people make allowance for all these things, it is much harder for software to do the same. That sort of software needs a very fast computer. The newest G5 Mac is a step in the right direction, it might be capable of running such advanced software. Mark- > > Like you, I was rather upset initially over the "poor quality" > > message but I gave it a chance anyway. > > I would be happy with "poor quality", however mine was "unacceptable" > so it would not allow me to continue the setup routine. If the mic doesn't work, then you can't possibly make a judgement about the software. Seems like replacing the mic would make sense before griping and sniping about software you haven't even tried, or declaring that the only good speech recognition software is made for "PC"s . Sounds like you have another, really obvious, agenda. <snip> > > I would be happy with "poor quality", however mine was "unacceptable" > > so it would not allow me to continue the setup routine. > > If the mic doesn't work, then you can't possibly make a judgement about > the software. Oh yes I can, and I am going to, loud and clear. Let's start with the name of the company that makes this crappy software, namely "MacSpeech, Inc." They have been making lousy speech app's for the Mac for years, but I thought I would give them another chance. Big Mistake. Apparently recently they have found it is a lot easier to sell defective software to PC users, because they don't even bother to include installation instructions for Macs in the pamphlet that ships with their iListen CD. Win 98, Win ME, Win 2000 yes, but no mention at all for OSX. They should change their name from MacSpeech to WinSpeech. Remember, this is the most _recent_ version of iListen, and MacSpeech Inc. does not list _anywhere_ what version of the Mac OS is required to run their software. The only reference I found to Mac OS X was this very misleading information about the Olympus DM-1 recorder that will transcribe the recorded speech directly into an OSX application of your choice, like "Word" for example: "These instructions show Mac OS 9 and Appleworks but the procedures are the same under Mac OS X, and you can transcribe into the application of your choice." That quote from an email reply they sent to me. It strongly indicates that iListen should work with Mac OSX "applications" - - - and of course OSX has to be running in order to "transcribe" the recorded speech to words in the running OSX application, Microsoft Word for example. Transcribing is a long process, iListen has to be up and running, "Word" has to be up and running, and of course the OSX itself has to be up and running. As the transcribing proceeds, under the direction of the iListen and Olympus-Recorder-software, you should be able to _see_ the text being typed into your OSX "Word" document at about 12 WPM (roughly, depends on the speed of your computer). Obviously, MacSpeech Inc. is making a deliberate attempt here to mislead prospective buyers into thinking that the latest version of iListen will work when installed into OS X. "the procedures are the same under Mac OS X" ...is the misleading passage in the email they sent to me when I was inquiring about their software. Well, they sucked me into wasting $150 on software that only runs on older obsolete Mac OS's and PCs, but it will be a cold day in hell before I recommend their software for anything except the trash can. > Seems like replacing the mic would make sense before > griping and sniping about software you haven't even tried...<snip>... You are the one who is quick to make knee-jerk judgements, not me. FYI, I have tried various incarnations of this crappy software over the years, and their latest version 1.6.4 is just as crappy as their previous versions were. > ...<snipped>...or declaring that the only good > speech recognition software is made for "PC"s . If you check this thread, I am not the only one who "declared" that. If you prefer to stick your head in the sand and believe that the crappy speech app's for the Mac are as good as the PC app's are, go right ahead. > Sounds like you have another, really obvious, agenda. Speak for yourself. It is quite obvious that your agenda is to demonize those who have different opinions than you do. Go away, I refuse to communicate with trolls like you who make accusations, instead of confining their remarks to the thread topic. Now that the troll "Madwen" is dispensed with, the rest of us can get back to discussing that lousy iListen software. In an attempt to discover just how lousy iListen is, I tried installing it on a Compaq Presario laptop, the very same laptop that I had installed a $600 version of "Naturally Speaking Pro" on. NP Pro is good speech software, even though the version I have is an older version, about five years old I think. I had no trouble installing it, no trouble training it, no trouble operating it. All-in-all a very impressive product, but very high in cost. Anyhow, back to iListen, created by MacSpeech, Inc. Inserted the iListen CD, it spun up but no icon appeared on the PC desktop. Bad Sign. The iListen installer app' appeared on the drive-E listing, but double-clicking failed to activate the install app', and "Open" also failed to activate the install app' Next, tried following the brief instructions on the included pamphlet for installing iListen on Windows ME. Supposedly by following those instructions two necessary drivers get installed. The OS went through the motions of installing the two drivers, but when I looked for the drivers later, they were nowhere to be found. Rebooted the PC. Tried again double-clicking the iListen install app', tried again using "Open" to start it, no luck, also still no icon on the desktop for the iListen CD. The pamphlet's written instructions gave no indication of what to do if the drivers refused to install. Next, I tried to use Window's "Restore" utility to restore the OS to its previous state, to undo any damage that the failed driver-install-attempt had done to the OS. "Restore" did not work, just displayed a message "Unable to Restore". Gadd, I Hate Windows! Advice from the general Window's books I was following for guidance: "Take your PC to a qualified Window's technician for installing drivers etc., because things like that are way beyond your capabilities as a user." They gotta be kidding, right? Mark- OK, Mark, Apparently you do have an agenda. I am not going to even attempt to reply to any of the so called accusations below except for one thing. MacSpeech does not make one piece of software that runs on a windows machine. We never have. iListen runs on Mac OS 9 and Mac OS X, and it runs like a champ. And our software does not come with a pamphlet. It comes with a CD. 100% of our documentation is on our CD. If you buy an Olympus Digital Recorder WITH iListen, you get what Olympus provides. We include instructions for using it within our user guide, but we had nothing to do with either the software or written material that Olympus includes with their recorders. I don't know why this person is trying to besmerch our company, but he just made a wrong turn. There is absolutely no way he could possibly even attempt to install iListen on a Compaq. It won't work. It's a Mac only CD! The installer will only run on a Mac! HELLO!??? Mark, please stop this. iListen works fine on Mac OS 9, and Mac OS X 10.1.5 and up, including Panther. It does not work on PCs. It never has. You are confusing the process of transcribing from the digital recorder to that of converting speech-to-text. Best Regards to all (except perhaps "Mark Conrad") Chuck Rogers, Chief Evangelist MacSpeech, Inc. > > > I would be happy with "poor quality", however mine was "unacceptable" > > > so it would not allow me to continue the setup routine. [quoted text clipped - 147 lines] > > Mark- > OK, Mark, > > Apparently you do have an agenda. I am not going to even attempt to > reply to any of the so called accusations below except for one thing. Oh, good lord, you're wasting time with Mark Conrad? Don't bother, Mark/Gerald is here specifically to waste the time of as many people as he can. > I don't know why this person is trying to besmerch our company, but he > just made a wrong turn. He's apparently taking a break from complaining that he can't partition his hard drive into sub-megabyte-sized partitions. Seriously. This, of course, is the fault of the "crappy disk manager application". Then there's the "Crappy Unix stuff that's so complicated and I don't want to read the man pages", ad nauseum. Nobody takes him seriously, so anything bad he says about your product is only going to serve as a reason for the rest of us to want to try it, rather than want to avoid it. Somewhere there is a "Mark Conrad FAQ" that a regular here has compiled, if you want more about why you can safely disregard the noisemaker, and why his badmouthing of any product isn't going to cause _anyone_ to seriously think it's true, google would find that FAQ. Best regards, Dave Hinz > OK, Mark, > > Apparently you do have an agenda. I am not going to even attempt to > reply to any of the so called accusations...<snipped>... Okay, perhaps I did come down too hard on MacSpeech, Inc., so I will give you the benefit of the doubt, temporarily. Statements like "Apparently you do have an agenda" and "Best Regards to all (except perhaps "Mark Conrad")" don't exactly get on my good side, though. > I am not going to even attempt to > reply to any of the so called accusations...<snipped>... Unlike you, I _will_ extend the courtesy to you of replying to your comments. > That VXI headset someone mentioned either was purchased from MacSpeech > or somewhere else. ViaVoice never shipped with that brand. I am the person who mentioned that headset. I also mentioned that I was not certain which speech app' the headset shipped with, and speculated that it was ViaVoice. Subsequently, I have found that _my_ VX1 headset actually shipped with Naturally Speaking Pro. FWIW, the "VX1 Parrott" headset is also sold seperately by MacSpeech. > They shipped only with either Andrea or Plantronics. The headset that MacSpeech shipped to me is a "Telex" headset, model number not specified either on the headset itself or in the pamphlet that came with the headset. > Also, the Griffin iMic has several revisioins. Older models will work > fine with iListen, while newer ones will not. Non-applicable here, I do not have a Griffin iMic. Chuck Rogers posted: "Our support department <questions@macspeech.com> is very good and will make sure you get up and running." When I double-click your link, I receive the error message: "The referenced article or group can not be found on the news server." > MacSpeech does not make one piece of software that runs on a windows > machine. We never have. Okay, I was wrong about my assumption that iListen would run on a PC. I was confused by the pamphlet that came with iListen's Telex headset, which had all sorts of setup instructions for various Windows OS's. It would be nice if MacSpeech would attach a disclaimer note to the Telex pamphlet to keep others from wasting their time trying to get iListen to work on a PC. > iListen runs on Mac OS 9 and Mac OS X, and it > runs like a champ. Did not run that way for me. > 100% of our documentation is on our CD. Not when you include the confusing pamphlet from Telex in your iListen package, without any obvious disclaimer about the PC sections of that pamphlet not applying to iListen. Any Mac user eager to get your product up and running might not spot the fact that the Telex pamphlet's info' about PCs has nothing to do with iListen. > I don't know why this person is trying to besmerch our company... Because your product did not work on my Pismo powerbook. I tried the Telex headset that MacSpeech shipped, and it did not even begin to work. I checked the obvious things, like the jacks being reversed on the Telex USB adapter. If your product _did_ work with my powerbook, I would be the first one to trumpet the benefits of your software. After all, I don't like the idea of having to use a PC just to get decent speech recognition performance. FWIW, when I finally got around to trying the different headset that came with Naturally Speaking Pro, then iListen performed without a hitch. iListen then recognized _all_ the initial training text. I did not have to repeat any word, not even one word. I am impressed. I have not got around to formal extensive training of your program to my voice yet, but will do so shortly. Confusingly, when I re-tried the Telex headset that MacSpeech shipped, this time it worked. Why, I don't know. As long as it works, I will continue to use the supplied Telex headset. One thing still puzzles me. In MacSpeech's ads, they trumpet the ability of iListen to do 'hands free' correction, or at least that is the way I understood the advertisement. That is one of the main reasons I bought iListen. Naturally Speaking Pro can do hands-free correction of mistakes, but so far I have not run across that feature in the iListen documentation. Can it or can't it do all corrections by voice only? (no keyboard needed) In any event, the entire correction window is non-intuitive. The PDF documentation does not make it clear how to get the 'correction' to be applied to correct the wrong word in my TextEdit app'. It also does not make it clear whether or not the correction gets applied to my voice profile, to prevent the same mistake from happening in the future. Naturally Speaking Pro is much more intuitive in their correction window terminology and labelling. I had no trouble at all in understanding exactly how their correction window worked. I intend to wring out your software, because I believe it can be adjusted to give almost the same quality of speech recognition performance as the much higher priced Naturally Speaking Pro. However, based upon my preliminary look at your PDF documentation, I don't have much faith in your documentation helping me much, and suspect I will have to learn how to use your product the hard way, by experimentation. > You are confusing the process of transcribing from the > digital recorder to that of converting speech-to-text. Nope, I know exactly the difference between the two. What you call "transcribing" I would more accurately call "moving" the WAV sound file from the recorder to the Mac, that way no one would become confused and think that your term "transcribing" would have anything at all to do with converting the WAV sound file to text. It does not, you know. One can move the sound file from the recorder to the hard drive, period, with no conversion to text taking place at all. All the conversion to text happens _after_ the entire sound file is moved to the hard drive. The conversion of the WAV sound file to text is a seperate long operation, which can be done at a later time if desired. At least that is the way it works with Naturally Speaking Pro, and I suspect it works the same way with iListen. BTW, thanks for being half-ways civil to me in this thread, and sticking to the thread topic instead of resorting to the usual character assasination that others in this NG inflict upon me. Mark- > Apparently recently they have found it is a lot easier to sell > defective software to PC users, because they don't even bother to [quoted text clipped - 4 lines] > > They should change their name from MacSpeech to WinSpeech. HUH???!!! Mark, you've really gone off into your own private dream world this time, because iListen is a Mac-only product. MacSpeech doesn't so much as mention Windows on their web site. SignatureMike Rosenberg <http://www.macconsult.com> Macintosh consulting services for NE Florida <http://bogart-tribute.net> Tribute to Humphrey Bogart Toyota Prius fans: Check out alt.autos.toyota.prius > > > I would be happy with "poor quality", however mine was "unacceptable" > > > so it would not allow me to continue the setup routine. [quoted text clipped - 3 lines] > > Oh yes I can, and I am going to, loud and clear. . ::laughs:: > Let's start with the name of the company that makes this crappy > software, namely "MacSpeech, Inc." [quoted text clipped - 10 lines] > > Win 98, Win ME, Win 2000 yes, but no mention at all for OSX. <snip relentless hornswoggle> You're obviously mega-certifiable. The product is made solely for the Mac. Reductio ad absurdum. Buhbye Markie... get help. . All: Let's see if I can weed through some of the comments made here, and shed some light on things. First, if you purchased iListen with a headset, there is absolutely no reason for it not to work unless something is defective, which does happen. We have thousands of users, and I just finished demonstrating iListen (very successfully) at an Apple store. Instead of complaining about the software, please contact <questions@macspeech.com> and determine what the problem is. We literally have 1000's of satisfied users, so your experience, while perhaps not unique, is not common either. Second, Mac OS X is very picky about audio - and it gets even pickier when you attempt to put text directly into a third-party app, the way iListen does. The Andrea headset someone mentioned is not supported by MacSpeech, although some of our users are using it successfully. That VXI headset someone mentioned either was purchased from MacSpeech or somewhere else. ViaVoice never shipped with that brand. They shipped only with either Andrea or Plantronics. Also, the Griffin iMic has several revisioins. Older models will work fine with iListen, while newer ones will not. Our support department can help you determine which one you have and whether you can use it with iListen. Our support department <questions@macspeech.com> is very good and will make sure you get up and running. Please try them before complaining to the world that our software doesn't work. You can check out some positive stuff people have said about iListen at <http://www.macspeech.com/news/testimonials.html>. Best Regards to all, Chuck Rogers Chief Evangelist, MacSpeech, Inc. > > > Like you, I was rather upset initially over the "poor quality" > > > message but I gave it a chance anyway. [quoted text clipped - 9 lines] > > <snip> > Just for kicks, thought I would evaluate MacSpeech's latest offering of > "iListen" version 1.6.4 for OS X. Ahh, this old chestnut. I hate to say it, but speech recognition on the Mac -- where it once led the field -- is now pretty useless, and as many speech advocates have discovered, if you want it to work, then you have to get a PC. There's even a doctor in here who specifically went out and bought a PC laptop to do dictation, but even that couldn't work reliably (*waves to Tim*) so now he uses a transcription service. He records into a simple audio recording app, encodes the notes into a mid-rate MP3, emails if off to his transcription service and gets the text file back within hours. One day, speech recognition systems will not only recognise every word you speak without interruption, but also be able to work contextually. I would *love* to be able to issue a command like: "Open the Freehand files from the Simpsons job. [computer opens last Simpsons file worked on] Mmm, no, the one from three days ago. [computer opens correct file]" Geoffrey (remove EXCESS BAGGAGE to reply via mail) SignatureWARNING: mail to this address will be auto-bounced if: (a) more than 10% original content appears before first quoted matter, (b) quoted material exceeds 75% of total message content, and/or (c) HTML is used to format text and/or embed non-ASCII items. > One day, speech recognition systems will not only recognise every word > you speak without interruption, but also be able to work contextually. Yep, the present offerings for both the PC and the Mac have yet to be able to recognize the context of a word in relation to other nearby words, phrases, and sentences. By far the best speech recognition software I have tried to date has been the PC-only software "Naturally Speaking Pro", sold by a firm called Lernout & Hauspie. Much better documentation about how to utilize their software, compared to the documentation and confusing window terminology from companies like MacSpeech. L & H in turn bought the software rights from "Dragon Systems", who developed the software for many years when it used to be called "Dragon Dictate". This topic of recognizing the "context" of a word in relation to other words and even other sentences is not an easy thing to do. It involves a huge software program with Artificial Intelligence components, and extremely fast computers, not to mention very high costs for the software itself. IMO, the only near-term solution we will be likely to see will be a seperate "correction program", which could be turned loose onto the present 98% average-recognition-accuracy files that are produced by today's software. The seperate correction program could take two inputs: 1) The written text file to be corrected. 2) The original WAV sound file. Overnight operation could keep the price of the correction program "tolerable" on today's computer hardware. The correction program, loaded with sophisticated AI components, could look at raw text like the following and correct the punctuation: that that is is that that is not is not ...by an involved process of looking for tell-tale pauses in the original sound recording, and observing what the context of surrounding words and sentences are, the automatic correction program could approach the level of a human and correct the punctuation and capitalization to correctly read: That that is, is; that that is not, is not. Homonyms and synonyms could be corrected also, by observing the context of words to their surroundings - - - stuff that historically has caused speech programs to stumble badly: ore scents sea red or sense see read (past tense of "read") our cents C oar ...also stuff like "chap an knees" versus "Japanese" Easy enough for a good AI program to spot and correct, in "context" that is. The reason the 98% accuracy rate is intolerable is because the necessary corrections slow down the whole process by an un-acceptable rate. Even a hunt-and-peck typist can beat the overall speed of the present speech programs, with better accuracy than the speech program offers. Now if the average accuracy went up from 98% to 99.8%, it would be a different story, and speech programs would be useful to many more people. Just my 2-cents worth, of course. Mark- All: First, I'd like to reiterate that we don't claim iListen to be as good as Dragon Naturally Speaking. They have had a lot more time to create a more mature product. We have only been at it since 1998, and then had to start basically from scratch for Mac OS X - so Dragon has a long head start. The good news is that we are continuing to improve the product, and for many users, it is more than just a toy. We have several doctors and lawyers and authors using the product on a daily basis. Next, I'd like to correct some factual errors in Mark's message: > > One day, speech recognition systems will not only recognise every word > > you speak without interruption, but also be able to work contextually. > > Yep, the present offerings for both the PC and the Mac have yet to be > able to recognize the context of a word in relation to other nearby > words, phrases, and sentences. This is not accurate at all. All present-day speech recognition programs, including Dragon, ViaVoice, iListen, Kurweil, and those made by the now defunct L&H use contextual recognition. Word recognition requires the user speak - in - distinct - pauses. All commercially available speech recognition programs today compare what was said with two or three words on either side of the current word being analyzed. But Mark's observations on the result is still essentially correct: 2 or 3 words, in most cases, does not constitute a phrase - but that is as much as today's computers can analyze in a reasonable amount of time in order to keep the latency between speech and typing to a minimum. As computers get faster, we can expect analysis to gradually extend beyond 2 or 3 words, which will dramaticlaly improve overall accuracy. > By far the best speech recognition software I have tried to date has > been the PC-only software "Naturally Speaking Pro", sold by a firm > called Lernout & Hauspie. Naturally Speaking Pro is currently sold by Scansoft, which bought L&H in a bankruptcy-forced sale due to their "Enron-style" management. The entire Naturally Speaking line of products is actually a part of Dragon, which was purchased by L&H. [snip] > IMO, the only near-term solution we will be likely to see will be a > seperate "correction program", which could be turned loose onto the [quoted text clipped - 20 lines] > > That that is, is; that that is not, is not. [etc...] Mark - you just described *EXACTLY* how the current correction process works in virtually ALL speech recognition programs currently available! Finally, for those of you who would like a "fair and balanced opinion," the most recent review of iListen was done by the Alaskan Macintosh User's Group. You can read it at <http://www.akappleug.org/rev/361.html>. Best Regards, Chuck Rogers, Chief Evangelist MacSpeech, Inc. Hi, Chuck. I vaguely remembered your name as the respondant to my inquiries when I tried to use iListen 1.0.1 way back in '98 or '99. I remembered that you signed "chief evangelist". I think it's great that you still are. I always like passion in people doing what they do! I'm impressed that you folks are as far along as you are. Dragon had only a 25-30 year headstart! And you guys are aiming at the niche market of Mac which itself is a niche market, so I am impressed with your courage and ambition. When I wrote with feedback, I was always impressed that you folks took my observations and criticisms seriously. For some reason one of the writers here says you folks ignore all criticism. Doubt that's true. Good luck. I'm a big fan/supporter of Mac platform. Hoping you guys succeed. Gotta keep going with Dell, Windows-based EMR (electronic medical record), and Dragon for now, but I'll keep checking in. Best wishes, -- Josh in Syracuse > All: > [quoted text clipped - 81 lines] > Chuck Rogers, Chief Evangelist > MacSpeech, Inc. <snip somewhat questionable "review" of iListen> > Most people's normal speaking rate is around 300 WPM. Does *anybody* find this claim credible? I don't. SignatureJim Glidewell My opinions only > > Most people's normal speaking rate is around 300 WPM. > > Does *anybody* find this claim credible? I think my sister-in-law probably speaks that fast, but no one has ever said she's normal. ;-) SignatureMike Rosenberg <http://www.macconsult.com> Macintosh consulting services for NE Florida <http://bogart-tribute.net> Tribute to Humphrey Bogart Toyota Prius fans: Check out alt.autos.toyota.prius > > > Most people's normal speaking rate is around 300 WPM. > > > > Does *anybody* find this claim credible? > > I think my sister-in-law probably speaks that fast, but no one has ever > said she's normal. ;-) heh heh, and a lot of TV announcers in commercial ad's seem to rattle on at that 300 WPM rate also. I had my doubts when I posted that figure of 300 WPM as being "normal", but I figured 'what the heck', if I am wrong, someone will correct me. Sure 'nuf one of my many critics did, but it took him long enough. I thought this thread was dead. <hrrump, cough, cough, gotta clear my throat now that the thread has sprung back to life again> Y'know, it kinda broke my heart when "iListen", the latest speech recognition product for OSX by MacSpeech, turned out to be such a second-rate product. (IMO, of course) Good side of iListen is that it only costs $99, so it is very cost effective for anyone who just wants to 'play' with speech app's. By comparison, my old version-5 of "Naturally Speaking Pro" for the PC cost me $700 when I bought it years ago. (medical and legal versions cost $1,000) That old version-5 beat the pants off iListen for the Mac, much better accuracy, much more 'consistant' as regards accuracy, far better correction system for correcting residual errors, and an excellent built-in tutorial that literally "teaches" one how to use all the features of the application. That excellent tutorial is worth the price alone. In a sexy female voice, it takes the user step-by-step through the application, demonstrating one small feature at a time, then allowing the 'student' to practice that feature until he gets it right. I was so impressed by the vast differences between iListen and the PC product that I shelled out $150 for an upgrade, which is now up to version 7.3 of Naturally-Speaking-Pro. Only down side is that I have to use PC hardware to run the app'. I really hate PCs, they seem so primitive compared to Macs. Mark- > Y'know, it kinda broke my heart when "iListen", the latest speech > recognition product for OSX by MacSpeech, turned out to be such a > second-rate product. (IMO, of course) Ditto opinion here. > I was so impressed by the vast differences between iListen and the PC > product that I shelled out $150 for an upgrade, which is now up to [quoted text clipped - 3 lines] > > I really hate PCs, they seem so primitive compared to Macs. Ain't it the truth. It broke my heart to buy a Dell laptop for the single solitary purpose of using speech recognition. But I had to because I wasted $99 on iListen 1.0.1 which they wouldn't refund when I gave up on it. Since then, I've used Dragon 3.5 (promising), Dragon 4 (solid), and Dragon medical version 5 ever since around 1999. It's been really good, not perfect, but really good. Excellent user interface. I've been so happy with version 5 that I haven't risked upgrading to 6 and 7! Now I'm considering it. I got the Dragon 7.3 medical upgrade for one of my partners who uses version 6. I gave it a try 2 days ago myself, 99.4% accuracy (wow! musta been a fluke) Nothing like it on the Mac platform. What a shame. -- Josh > > I really hate PCs, they seem so primitive compared to Macs. > > Ain't it the truth. It broke my heart to buy a Dell laptop for the > single solitary purpose of using speech recognition. Sorry for the delay. I have been busy lately trying to talk my surgeons into installing a zipper in my chest. <long story> Yeah, I know what you mean about the Dell laptop. I have three old PC laptops that I occassionally press into action, mostly for speech recognition projects. > ...<snipped>...because I wasted $99 on iListen 1.0.1 which they > wouldn't refund when I gave up on it. I did not even try to get a refund. I was fairly certain that iListen could not cut the mustard, however I like to keep my eyes on them every year or so, just in case they get some smarts and start listening to constructive criticism from potential new customers who have been using the PC version for years. Lately, they seem more interested in defending their 2nd-rate product than they do in changing it to make it more like the PC application in real performance features. Heck, even the much better PC app' is not yet suitable for the majority of things that one might use a speech-recognition program for. > Since then, I've used Dragon 3.5 (promising), Dragon 4 (solid), > and Dragon medical version 5 ever since around 1999. It's > been really good, not perfect, but really good. > Excellent user interface. I used the very old "Dragon Dictate" back in the days when the best it could do was 40 WPM, and each word _had_ to be spoken seperately with slight pauses between words. Many years later, I used ver' 5 of Naturally Speaking Pro for transcribing my close friend's crude notes into the first drafts of the three books that he was authoring. He was a UPI war correspondent in the Pacific, got a Purple Heart medal there, even though he was a civilian. Lotsa Japanese names, repeated in profusion throughout his books, so using the 'spell' feature slowed me down too much. I eventually had to create a macro to modify the built-in vocabulary to accomadate the repeated use of those names, military terms, and Japanese island names. Fortunately, version 5 supports _quick_ generation of macros that modify the main vocabulary of the speech program. > I've been so happy with version 5 that I haven't risked upgrading to 6 > and 7! Now I'm considering it. Same here, version-5 is a winner. I recently bought the ver' 7.3 upgrade, because I fell victim to their "improved accuracy" claim, but have not got around to installing it yet. > I got the Dragon 7.3 medical upgrade for one of my partners > who uses version 6. I gave it a try 2 days ago myself, > 99.4% accuracy (wow! musta been a fluke) Glad to hear that. > Nothing like it on the Mac platform. What a shame. Yep. Well, we can always hope that MacSpeech will eventually get out of defensive mode and start listening to constructive criticism. The speech recognition market is a tiny market segment, I would hate to be competing there. If MacSpeech decides to invest the money needed to become a main player in that small market, it is a huge risk for them. Just for beginners, they would have to drastically increase the price of iListen, to somewhere around $500. This would be needed to attract the kinds of programmers who could revamp iListen to be competitive to the PC version. That revamp could take some time. Among other things, they would have to worry about software patent infrigement, when the PC guys detect that they are trying to become a real players in this limited market. Success in the speech recognition market involves creating software that pays close attention to the following factors: 1) Accuracy, Accuracy, Accuracy. So far, no speech recognition software has the required accuracy; even the PC versions are lacking in the required accuracy. Slight gains in initial accuracy attract customers, because each gain in _initial_ accuracy means that speech recognition becomes more suitable for practical application to real-world uses. 2) A quick, easy, and 'intuitive' correction system for correcting residual errors. It does little good to be able to dictate at 160-WPM if the resulting 160 WPM is lowered to 30 WPM or worse by the time lost in correcting mistakes. When all is said and done, a skilled operator can only correct a few mistakes a minute, even using the good PC programs which have a much better correction system than iListen has. To complicate this matter of 'correction', the correction chore is often delegated to subordinates, because people like doctors just do not have the time to correct their own initial dictation. It could really screw things up, to say the least, when a mistake like "fifty milligrams" was intended to be "fifteen milligrams". Some people tend to "trail off" the "n" sound in the above example, which makes it very hard for the program to tell the difference between "fifty" and "fifteen". A 'responsible' doctor would likely make time to have the computer 'read-back' the _corrected_ dictation in a computer-generated voice, in order to detect any life-threatening errors. Even a busy doctor should make time for that, for obvious reasons. 3) A built-in training system for how to use the program to obtain the greatest possible initial accuracy. I am not refering to a users manual, but rather a step-by-step tutorial program with user participation in each step by actually trying to implement the step in an example window which is part of the tutorial program, like the PC application "Naturally-Speaking-Pro" has. The Mac iListen program has no such step-by-step tutorial. For the distant future when computers become much faster, and speech-recognition applications incorporate advanced Artificial-Intelligence techniques, there is the distinct likelyhood that life-threatening errors like mistaking "fifteen milligrams" as "fifty-milligrams" could be caught by the program itself. The program could be smart enough to 'recognize' the erroneous dosage as being a obvious error, based upon the ordinary dosage for the affliction, which could be in the program's database. Mark- > That revamp could take some time. Among other things, they would have > to worry about software patent infrigement, when the PC guys detect > that they are trying to become a real players in this limited market. Patent fights on SR go back at least to the mod-80's (between Dragon and IBM, in my fuzzy recollection). > 1) Accuracy, Accuracy, Accuracy. > So far, no speech recognition software has the required accuracy; > even the PC versions are lacking in the required accuracy. > Slight gains in initial accuracy attract customers, because each > gain in _initial_ accuracy means that speech recognition becomes more > suitable for practical application to real-world uses. The criteria for court stenographers (who are the highest skilled human transcriptionists) is about 200 wpm at 95% accuracy. I google'd some job listings for shorthand stenogrpahers for government office work where the requirement was 35 wpm at 95% accuracy. I bet that medical transcriptionists used by most hospitals to handle physician taped dictataions would not be much better than 95%. Especially when you note similarity of sound of similar drug names (Digoxin, Digitoxin, Doxepin) and common typos missed by spell checkers (their/there its/it's etc) Plus, if you dictate a letter to a voice recorder and play it back, I bet you find several "speechos" (spoken typos). A skilled personal secretary who has worked for the same person for a while usually will automatically make appropriate minor grammatical corrections when transcribing, knowing the boss's style. An anoymous transriber will likely make no assumptions about what the speaker meant, but transribe verbatim, appearing as an error when the speaker reads it back. > The program could be smart enough to 'recognize' the erroneous dosage > as being a obvious error, based upon the ordinary dosage for the > affliction, which could be in the program's database. This is the "Andromeda" model - a computer which is smarter than you and will always do what you MEAN and not what you SAY. As opposed to the "Star Trek" model, where the ship computer was so dumb, they needed wetware to aim the weapons and plot the course and you had to be careful about it crashing/freezing due to logical contradictions ("Computer, everything I say to you is a lie"). > > The program could be smart enough to 'recognize' the erroneous dosage > > as being a obvious error, based upon the ordinary dosage for the > > affliction, which could be in the program's database. > > This is the "Andromeda" model - a computer which is smarter than you and > will always do what you MEAN and not what you SAY. Yeah, kinda. What I had in mind was a computer that would check in with me after it proofread my dictation and say something like: "Did you _really_ mean to prescribe 50 milligrams of Lipitor, instead of the more commonly prescribed 15 milligrams?" A "back seat driver" sort of correction program.<g> Oh, before any heart patients out there get too excited, I realize that the really "normal" daily dosage for Lipitor is generally 10 milligrams. > I bet that medical transcriptionists used by most hospitals to handle > physician taped dictataions would not be much better than 95%. > Especially when you note similarity of sound of similar drug names > (Digoxin, Digitoxin, Doxepin)...<clipped>... Probably... That is why it is so important that a doctor take the time to thoroughly review his "corrected" transcription, in order to root out any remaining mistakes. > Plus, if you dictate a letter to a voice recorder and play it back, I > bet you find several "speechos" (spoken typos). Tell me about it! I caught myself making mistakes like that. I really like my older Sony recorder model ICD-MS1. It is so small that fits in my shirt pocket with nothing sticking out. Using its built-in mic', I have had runs of over a hundred words with no mistakes at all. I hope the NSP upgrade ver' 7.3 still supports that older recorder, the documentation does not make it clear whether it does or not. The newer Sony recorders are bigger, darn it. Present day speech programs really do their best to "guess" what our utterings mean. Just for kicks, I cleared my throat when hooked up to iListen, and it promptly ground out a whole string of words. Tried it again, and a whole bunch of different words came out. Up until then, I did not realize that my throat-clearing was so expressive. I will hafta be careful when speaking to others, not to clear my throat, else they might take offense at what they thought I said. Nonsense noises should not result in output from speech programs. Even Naturally Speaking Pro is guilty of this flaw, although to a much lesser extent than iListen. I shelled out close to $1,000 so far on the PC version. I am in accord with others here, in that I would prefer to be spending all that money on iListen instead, when their software becomes more competitive. Think I will wait a bit before installing the 7.3 upgrade for Naturally-Speaking-Pro. I am nervous about windoze possibly zapping my extensively trained voice files. I have an old "Norton's Ghost" util' to backup the PC, but Ghost is far from being a GUI utility, very non-intuitive to use with all tha arcane DOS command-line switches. Almost as bad as UNIX. Think I will check Google to see if a more recent GUI version of Ghost is available. Wanna make an image-type of backup file of the "C" partition of the PC, one that will result in a bootable C partition after I restore. Mark- >>That revamp could take some time. Among other things, they would have >>to worry about software patent infrigement, when the PC guys detect [quoted text clipped - 23 lines] > (Digoxin, Digitoxin, Doxepin) and common typos missed by spell checkers > (their/there its/it's etc) I'm happy to chime in and give 2 responses: 1) hospital/office medical transcriptionists are WAY better than 95% accurate because they do know what the doc is trying to say and they do use spell-checkers to catch typos. Honestly, I can't remember the last time I saw a typo in my hospital dictation transcriptions. When the transcriptionist is stumped, they don't guess, they leave a blank space with an underline so the doc can fill in the correct word. 2) doctors generally do NOT check their transcriptions! I know, gasp, shriek. But no one has time to check a 1000 word discharge summary for its possible 1 or 2 typos which have a vanishingly small likelihood of changing the meaning of what one dictated. Yes, that opens the door a tiny crack for medical errors or for lawyers to tear us apart, but only a microscopic crack. Because it is true, transcriptionists almost never make an error (maybe 1 out of 1000 words wrong), and among those errors, almost none (perhaps another 1 out of 100) changes the meaning of what was written. Multiply those together and you are looking at careful reading of 100,000 documents just to find one error, and that's obviously never going to happen. The final screen to catch errors is that any doctor reading that document a day, month, or year later then has to think about the info in that document. If it doesn't make sense, don't trust it. We don't just blindly take as truth whatever is contained in any document, any history, any lab report, etc. Human thoughtful consideration is the final screen to catch errors, and it's pretty good too. Putting on my bicycle helmet to protect myself from anyone who wants to bash me with the Institute of Medicine's (IOM's) Report on Medical Errors, -- Josh Steinberg MD > Putting on my bicycle helmet to protect myself from anyone who wants to > bash me with the Institute of Medicine's (IOM's) Report on Medical Errors, > > -- Josh Steinberg MD heh heh, no one here is going to bash you for that enlightening post. That certainly eases my mind about the 95% accuracy criteria for medical transcriptionists - - - that in practice they are much better than that because of the close cooperation with the doctors. > ... doctors generally do NOT check their transcriptions! I know, gasp, > shriek. But no one has time to check a 1000 word discharge summary > for its possible 1 or 2 typos which have a vanishingly small > likelihood of changing the meaning of what one dictated. I would imagine that when a doctor hires a _new_ transcriptionist that he takes the time to check the work, until he is confident about the skill level of that new person. ...if for no other reason than to keep his malpractice insurance premiums down to their modest $15,000 a year cost.<g> Every doctor I have run across is busier than a one-eyed cat in a fish market, which is why they have to delegate work to the hired help. Mark- Mark: This simply isn't true. We take every criticism of iListen very seriously, and incorporate all into our overall product plan. Just because things you don't see a feature appear, doesn't mean we don't know about it or don't care to implement it. For one thing, we had to completely rewrite the program from scratch to work with Mac OS X. (If you remember, Dragon announced they were going to support Mac OS X and then gave up when they found out how hard it was). Developing complex software takes two things: time and money. The less money you have, the more time it takes to implement people because the fewer man-hours you have available. If you know of anyone who would like to join our investor ranks with a significant influx of cash, I am sure we could get some of the features you are looking for implemented much more quickly. Unless you are, however, you will have to put up with what we can provide when we can provide it. At the very least, please stop using terms like "2nd rate product." It is a first-rate Macintosh product and always has been. Just because a program doesn't measure up to the 800 lb. gorilla in the marketplace, doesn't mean it is "2nd rate." That implies a company that is not interested in developing first-rate software. I assure you, MacSpeech is interested first in making what we have work, then adding new features. I really don't see how that makes us or our software second rate. Yes, I am defending the software and the company with the above statement. That's my job. That doesn't mean we are more interested in that than in developing quality software, however. It just means that we need to respond to certain individuals who would put us in a bad light. Best Regards, Chuck Rogers, Chief Evangelist MacSpeech, Inc. Mark Conrad <NoSpamDammit@invalid.com> wrote in message news: [snip] > Lately, they seem more interested in defending their 2nd-rate product > than they do in changing it to make it more like the PC application in > real performance features. [snip] > At the very least, please stop using terms like "2nd rate product." It > is a first-rate Macintosh product and always has been. Well, when most people rate products for a particular purpose, their best product is often called a "first rated product", and lesser products are often called "second rated", "third rated", etc. That is why people "rate" products, to determine which products are the best, and which products are better than other products, i.e. their "rank" or "rating" in the hierarchy of best, average, worst, etc. One would think you would be happy with "2nd rate product", that puts you ahead of others like ViaVoice. Rating a product has absolutely nothing to do with the motivation of the people who create that product, as you stated erroneously in your post. People can be highly motivated and _still_ create a 2nd rate product. People do _not_ rate products by the motivation of the company that created the product, instead they rate products by how well they perform the task under consideration, the task they are being rated for. I and others care less about how much a company is motivated, that does not feed the bulldog. I am in no way faulting MacSpeech Inc. for fielding a 2nd rate product, they are doing the best they can given their available resources, and they are to be commended for that. I and others are looking forward to the day when MacSpeech creates a 1st-rate-product for converting speech dictation to text. Nothing would make me happier than seeing Naturally Speaking Pro relegated to 2nd rate status, well behind iListen by MacSpeech. I wasted all day yesterday fighting with PC software in my futile attempt to try to backup a PC partition. I kept getting obscene messages about not having this or that driver, and other messages about buffer overruns and needing to import Microsoft "patches" to prevent such overruns, but the straw that broke my back was when the XP operating system itself arbitrarily decided to shut my computer down "for my own good", or words to that effect. I completely reloaded the entire damn XP OS from its CDs, several times, each time taking about 30 minutes, but no luck in throwing the myriad faults. Even tried running diagnostics, but hardware passed ok. All I was trying to do was simply to download and install a needed "Framework" file named ".NET Framework 1.1" which was needed before I could install the latest version 9.0 of "Norton's Ghost", an image backup/restore utility capable of restoring a bootable drive C on my PC. You think I like doing this crap? Get yourself busy creating something for the Mac that comes closer to being a 1st rated product, please! As for my PC, I _need_ that image backup utility, in order to protect my extensive training files for speech recognition on the PC, not to mention all the various drivers and security patches, and the miserable PC "Directory", which apparently gets modified behind my back by every program in creation. (Welcome to the world of PCs, gentle people) I am going to give up and do what other PC users are forced to do, namely take the PC to a shop and pay them big bucks to install the needed files. grrr Mark- > > At the very least, please stop using terms like "2nd rate product." It > > is a first-rate Macintosh product and always has been. [quoted text clipped - 9 lines] > One would think you would be happy with "2nd rate product", that puts > you ahead of others like ViaVoice. Kinda like the fact that you should be grateful if we refer to you as a "second rate poster". Or even third rate, for that matter... SignatureJim Glidewell My opinions only > > One would think you would be happy with "2nd rate product", that puts > > you ahead of others like ViaVoice. [quoted text clipped - 3 lines] > > Or even third rate, for that matter... Hey, there are a lot of very smart people posting in this NG, who I would rate well above myself. I would be deliriously happy if you refered to me as a "75th-rate-poster" for example. In that case I would know that I was really making progress with this Euunix crapola. Mark- > As for my PC, I _need_ that image backup utility, in order to protect > my extensive training files for speech recognition on the PC, not to [quoted text clipped - 7 lines] > namely take the PC to a shop and pay them big bucks to install the > needed files. grrr Well the PC store people are really sinking their hooks into me, I never should have told them I was a Mac guy. Their original estimate for downloading and installing the needed free PC-file named ".NET Framework 1.1" was $30, and they further said it was no big deal and should only take an hour to do. That free file is needed in order for the newest "Norton's Ghost" backup program to work, which Symantec corp' mentions in small print as one of the 'requirements'. After I took my PC laptop model-5425 into their shop, their estimate now has climbed to $60 for installing one crummy free file. (they said their estimate might run even higher later) Those PC 'experts' had as much trouble trying to download that free file as I had encountered. They now say they will need my PC for at least another day, at $55 an hour for labor costs. It is amazing to me that the PC users in this area put up with that sort of stuff. They apparently get sucked in by the low initial price of a PC, then later really pay the piper when they find that they are not up to doing the myriad things that are necessary to keep their PC going. (like my present problem of downloading and installing the free library file) Lots of this stuff that is 'necessary' on a PC is fully automatic on a Mac, such as installing drivers, installing needed software libraries, modifying the PC "Registry" file, etc., etc. On the surface, it appears that the various "wizard" software stuff can solve PC problems, but often those wizard programs do not cover specific problems. When that happens, the clueless PC users have to pay the high priced 'experts' to get things running. That is why there are so many PC shops in my area of California, all of them engaged in a feeding frenzy off the clueless PC users here. Not one Mac shop here in 40,000 square miles, the Macs just, er, "keep-going" with no problems. The only real Mac problems I had in years were related to failed hardware. One problem was a failed memory-manager chip on a new Mac, which was fixed by Apple at no cost to me, not even shipping costs. Another problem was a bad keyboard on a really old Mac powerbook. I had to drive hundreds of miles to even find a Mac repair shop. They found the problem and installed a new keyboard at a very modest cost. I shudder to think how much it would have cost if I dealt with the PC shops around here. Mark- So, Mark - by the definition you present below, it is safe to say you believe we are a first-rate Macintosh speech recognition product, then? I am sorry to pick-nits, but when most people (other than you, apparently) says "second-rate" it is not a compliment. So I would really rather we didn't go around promoting the fact we are "second-rate." According to thefreedictionary.com, "second-rate" means "moderate to inferior in quality." It's that "inferior" part we are trying to avoid, because the vast majority of our users certainly wouldn't classify us as that. At this point, we aren't comparing ourselves to Dragon Naturally Speaking. We know it is better - but they had about a 20 year head start on us. Give us some time to catch up. I think it is safe to say that in the time since we released version 1.0 we have made tremendous strides in quality, accuracy, stability, and features. This is despite the fact that we had to practically rewrite the software from scratch to support Mac OS X. All I am asking here is that you take a more optimistic "glass is half full" approach in your assessments. There is nothing wrong with saying we don't compare with Naturally Speaking - and we are the first to admit that. But what is wrong with saying something like "nonetheless, they are the best available on the Mac and doing a great job of constantly improving their product," which is exactly what we believe we have been doing. BTW, it is totally understandable for you and others to be impatient for iListen to mature more rapidly. Lots of coding requires lots of time. The only way to cut the time to deployment is to throw more people at the problem, which costs $$. You would be amazed at how quickly we could improve the product if we could only raise a fraction of the investment income raised by Google. Heck, if each of the people reading this message got just 4 or 5 people to buy the product, it would speed things up considerably! So, when you see fit to complain about what iListen lacks - make sure you also complain about the lack of proper investment, and the lack of willingness of tech investors (and others out there) to put money into a company that is diligent, dedicated, and has a proven track record. Best Regards, Chuck Rogers, Chief Evangelist MacSpeech, Inc. > > At the very least, please stop using terms like "2nd rate product." It > > is a first-rate Macintosh product and always has been. [quoted text clipped - 70 lines] > > Mark- > So, Mark - by the definition you present below, it is safe to say you > believe we are a first-rate Macintosh speech recognition product, > then? Yes, as long as we all do not rate your product against the PC software, then we can safely say that your MacSpeech software is indeed a first-rate Macintosh speech recognition product. > Heck, if each of the people reading this message got just 4 or 5 > people to buy the product, it would speed things up considerably! I made my small contribution of $99 recently, even though I was fully aware that the iListen software could not compete head-to-head against the PC software. On the positive side, if a Mac user wants to see the present status of speech-to-text software on the Mac, I would not hesitate to recommend your product, which only costs $99. The PC alternative is very expensive and difficult to implement: 1) Buy a fairly recent PC machine and an external drive for it. Make sure it uses the XP operating system, primarily because the recent version 9.0 of "Norton's Ghost" backup software will not run on a lot of other PC OS's, which is why XP is needed. Older DOS versions of Norton's Ghost are very difficult for a Mac user to master, due to their arcane DOS oriented switches. 2) Install all the security patches _before_ attempting to use the PC. 3) Download the very large free file ".NET Framework 1.1" which is almost impossible to find on the Symantec website. Install that file in the appropriate place on the PC. That file is necessary in order to get Symantec's "Norton's-Ghost 9.0" to run, which in turn is needed to backup the C partition of the PC, in order to protect the valuable voice training files. 4) Buy "Norton's Ghost" ver' 9.0 for backup/restore. 5) Purchase "Naturally Speaking Pro" version 7.3 which costs around $700. In my case, I also had to shell out $83 to my local PC shop just in order to overcome the difficulties in getting that big "Framework" file downloaded and installed. In the past, I also bought the Sony digital recorder model ICD-MS1 which cost an additional $300. I used it all by itself, (no computer) to record my voice from reading hundreds of scrawled handwritten notes. My recording was used to initiate the initial drafts for 3 books about WW2 in the Pacific by my best friend who was a UPI war correspondent there. He died recently at age 89. Later, the recording was dumped into the PC, where it was automatically converted to text. Chuck, keep up the good work. A lot of us hate using the PC so we are watching the progress being made by MacSpeech with interest. Mark- Is there a way to use Naturally Speaking on a Mac with something like Virtual PC? I am expecting a new OS X, G4 powerbook this Wed, and one of the very wonderful things i was looking forward to was the ability to finally use voice dictation. (my cache-less wallstreet is too slow for any such software.) Must I get a PC to use NS? thanks, cathy In article <674c19c39d0ca4747d6803bf118a2a28@localhost.talkaboutmac.com>, > Must I get a PC to use NS? I'm new to (attempts at) VR. What's the deal with iListen? Sorry, I didn't see the earlier part of this thread. Is iListen a viable OSX VR application? Thanks, RB > Is there a way to use Naturally Speaking on a Mac with something like > Virtual PC? No. Don't even think about it. I am expecting a new OS X, G4 powerbook this Wed, and one of > the very wonderful things i was looking forward to was the ability to > finally use voice dictation You can: you can use iListen or ViaVoice. Both work fine (and they work better the faster your computer is and the more memory it has). m. Signaturematt neuburg, phd = matt@tidbits.com, http://www.tidbits.com/matt/ AppleScript: The Definitive Guide http://www.amazon.com/exec/obidos/ASIN/0596005571/somethingsbymatt Read TidBITS! It's free and smart. http://www.tidbits.com catgil: > > Is there a way to use Naturally Speaking on a Mac with something like > > Virtual PC? matt neuburg: > No. Don't even think about it. > > I am expecting a new OS X, G4 powerbook this Wed, and one of > > the very wonderful things i was looking forward to was the ability to > > finally use voice dictation > You can: you can use iListen or ViaVoice. Both work fine (and they work > better the faster your computer is and the more memory it has). m. You may be living in a reverse parallel universe. ViaVoice doesn't work at all -- it stops accepting dictation after a period. And that's not only my experience. iListen has potential, but has not arrived yet. The answer is that for a good voice recognition experience one needs to run Dragon Naturally Speaking on a PeeCee. Even ViaVoice on a PeeCee is better than anything on a Mac. Sorry to have to say this. Davoud > catgil: > > > Is there a way to use Naturally Speaking on a Mac with something like [quoted text clipped - 12 lines] > You may be living in a reverse parallel universe. ViaVoice doesn't work > at all -- it stops accepting dictation after a period. It is perfectly possible that I am living in a reverse parallel universe. The words you are reading now were entered by speaking into a microphone using ViaVoice. That includes the punctuation. You will notice that the program did not "Stop accepting dictation". m. Signaturematt neuburg, phd = matt@tidbits.com, http://www.tidbits.com/matt/ AppleScript: The Definitive Guide http://www.amazon.com/exec/obidos/ASIN/0596005571/somethingsbymatt Read TidBITS! It's free and smart. http://www.tidbits.com > The words you are reading now were entered by speaking into a > microphone using ViaVoice. Please, any info on ViaVoice appreciated. It's my understanding that VV doesn't exist anymore for the Mac - is that true? If so, what (older) version are you using, and on what kind of OS, etc? I have a new G5 dual with lots of RAM and want to get VR going.... Thanks, RB > > The words you are reading now were entered by speaking into a > > microphone using ViaVoice. > > Please, any info on ViaVoice appreciated. It's my understanding that VV > doesn't exist anymore for the Mac - is that true? If so, what (older) > version are you using, and on what kind of OS, etc? I use the latest version of the Panther. VV is no longer supported (developed) but it is still sold. iListen is both supported and sold. m. Signaturematt neuburg, phd = matt@tidbits.com, http://www.tidbits.com/matt/ AppleScript: The Definitive Guide http://www.amazon.com/exec/obidos/ASIN/0596005571/somethingsbymatt Read TidBITS! It's free and smart. http://www.tidbits.com In article <674c19c39d0ca4747d6803bf118a2a28@localhost.talkaboutmac.com>, catgil > Is there a way to use Naturally Speaking on a Mac with something like > Virtual PC? Unfortunately, no. VPC can not use any sort of microphone input from the Mac side. I have the same problem here. I run the latest version of VPC on my Mac Pismo, and I also have Dragon NaturallySpeakingPro version 8.0 on an old PC, a HP N5425 that I keep for that app' only. The PC runs the XP (home-version) OS. I am strictly a Mac person, having used them since the original Mac came out in 1984, and used Apple computers before that since the days of the Apple-2 computer. ...but I draw the line on using any kind of Mac-compatable dictation software, it is years behind Dragon's PC software, accuracy wise. Voice recognition technology is one field where a slight added amount of accuracy in recognition literally costs an arm and a leg. The PC stuff is very expensive. I have shelled out over $1,000 for dictation software and hardware, but it is the only way to fly when one is serious about dictation applications. The "Pro" version alone costs $700, but it has enough added features that I bought it instead of the much lower priced "Prefered" version, which I think runs about $200 or slightly less. Mark- > I am strictly a Mac person, (blah blah blah)... ...so you see I have no axe to grind about PCs being inherently 'better' at voice-recognition, or any nonsense to that effect. Voice recognition software is a small niche market, so the developers who got into that business originally chose the PC, because that would provide them with more potential customers. One of the first companies sold software called "Dragon Dictate", which did a fairly good job if you spoke slowly at 40wpm or less, with distinct pauses between words. That company evolved through many different owners, the present owners being "Scansoft.inc.", and the product name is now "Dragon NaturallySpeaking" which is sold in many different versions. The speech recognition "engine" in all the various versions is basically the same, from the low end "Preferred" version ($186)up to the high-end $1,000 versions, so the basic recognition accuracy and speed is about the same between all their versions. However, like CPU speeds in computers, there is a heck of a lot more to voice-recognition software than the basic recognition "engine". There are $1,000 versions for the Legal and Medical professions. They can now handle continuous-speech at about 200wpm with a claimed recognition accuracy of 99% if the user takes the necessary time to "train" the software, a somewhat long and tedious process. Even the best of these well-trained programs do not have anywhere near the recognition accuracy of a normal human. Despite that, at times I have runs of 200 words at a time without any mistakes, other than punctuation. ...but two hundred words without a mistake is only one minute of fairly fast speaking<g> Low-budget medical organizations can not always afford a team of human specialists to document a tricky operation, so they make do with a few camcorders placed in stategic locations, and all the surgeons wearing noise-cancelling wireless headsets that transmit directly to nearby PCs where their running accounts are changed directly to text. The high-cost "Pro" version of "Dragon NaturallySpeaking" saves the original audio from each surgeon's headset as a "WAV" sound file, to be used later to (painfully) correct the errors in recognition that always occur. The low-cost "Prefered" version does not save the original audio. FWIW, which isn't much, the correction process itself can be done entirely by voice, no keyboard required, your hands behind your neck. Some day I will think up a useful reason to use that "feature". I hate correcting the text, it is a slow tedious process. I have had to correct text for books that have many thousands of words, so the painful correction process gets old very quickly. THAT is the main reason I refuse to use iListen, ViaVoice, MacSpeech, etc., etc. - - - and instead use the PC software/hardware, to side-step most of the gawd-awful correction chores. If someone created software that increased average recognition accuracy from 99.0% to 99.5% i would shell out hundreds of extra dollars to buy that software. That is not likely to happen in our lifetimes, though. Mark- Thanks Mark and Rob for real world / detailed info. And all others for info as well. I may look for a cheap 2nd hand pc for dictation.. Is there a way to hook up a pc to a mac to transfer files (ethernet)? Or will i end up using a floppy a lot? cathy catgil: > I may look for a cheap 2nd hand pc for dictation.. Is there a way to hook > up a pc to a mac to transfer files (ethernet)? Or will i end up using a > floppy a lot? You can network Macs and PeeCee's via Ethernet or Wireless. What's a "floppy?" Davoud :catgil: :> I may look for a cheap 2nd hand pc for dictation.. Is there a way to hook [quoted text clipped - 3 lines] :You can network Macs and PeeCee's via Ethernet or Wireless. What's a :"floppy?" It' truly simple, too. 10.2.8, and probably Panther, ships with Samba (which identifies itself as Windows NT 4.5). You just turn on Windows file sharing, and all those Mac volumes just pop up on your PC "desktop." If you don't want to do that, you can use the built-in ftp server that comes with the Win NT series (I know nothing about Win 95/98/ME). Then you can move files back and forth that way. I recommend NCFTP for your Mac client. Like the man said, "what's a 'floppy?'" There haven't been floppy drives on Macs for a couple-three years now. SignatureLooks like more of Texas to Me In article <ccaf6bb4cde69edaa55e408397628bdb@localhost.talkaboutmac.com>, catgil > I may look for a cheap 2nd hand pc for dictation. Not too cheap.<g> Remember that Dragon is a resource hog, they recommend 500MBs of RAM for the Pro' version. If you have the "Preferred" version 7, now might be a good time for you to upgrade to the more accurate Preferred 8.0 version, $150. They don't give you much of a price break on the "upgrade", because they realize that they are the only game in town, accuracy wise. I think you only save about $50. It might be better just to buy the new Preferred version 8.0 outright, instead of the "upgrade". That way you don't have to suffer through two seperate installs. To "Activate" the new version 8.0, you have to connect your PC to the Internet.<shudder> ...otherwise, the new version will stop working after 5 times. In my case, I had to shell out $600 for the 8.0 upgrade, because I have the Professional version. Out and out purchase of the 8.0 Pro' version is $700, so I only saved $100 by upgrading. I always buy the CD, instead of the electronic download. For any lurkers who might go to the website for Dragon, beware to _not_ click on the "Mac version" link there, because otherwise you will be routed to IBM's "ViaVoice", which is an entirely different inferior product. <http://www.scansoft.com/> (follow the "Dragon NaturallySpeaking" link) Like I implied in previous posts, the added accuracy of Dragon pays for itself over and over, in the time and aggravation saved in the "correction" process. Mark- In article <ccaf6bb4cde69edaa55e408397628bdb@localhost.talkaboutmac.com>, catgil > Thanks Mark and Rob for real world / detailed info. And all others for info > as well. I shudder when I see the words Mark and real world in the same sentence. I use iListen continuously and I find it to be a very good solution. > I may look for a cheap 2nd hand pc for dictation.. Is there a way to hook > up a pc to a mac to transfer files (ethernet)? Or will i end up using a > floppy a lot? > > cathy > Is there a way to use Naturally Speaking on a Mac with something like > Virtual PC? I am expecting a new OS X, G4 powerbook this Wed, and one of [quoted text clipped - 3 lines] > thanks, > cathy Don't bother to get NS on the off chance it will run adequately on VPC. Dictation software needs lots of fast processing power. VPC optimised on a G4 runs at what I estmate to be a notional processor power of - say - 250-300mhz. It's just adequate to play music files without clicks and pops, let alone run dictation software. Personally I have made peace with Via Voice for Mac OSX and although it's full of bugs, it works adequately. As recognition software I find it fundamentally very good, it's just that it was released with loads of infuriating bugs and there's never been any further development of it. You can get it for a song. Rob >>Most people's normal speaking rate is around 300 WPM. >> > Does *anybody* find this claim credible? It is not true. Human speech is appx 120-150 words per minute. Try it yourself. Grab your local Sunday paper (or any online since you are sitting at your computer). Get out a watch or timer with a second hand. Start reading the lead story in the newspaper at normal speech pace. When 60 seconds is up, go back and count how many words you read. You'll see. It ain't 300. -- Josh |
Enable EMail Alerts
Start New Thread
Reply to this Message