Talk:MacSpeech Dictate

Lacks Key features found in NatSpeak for Windows
The key learning feature is lacking, and until it is included, the software cannot be trained.

Reviewing MacSpeech Dictate in the New York Times, David Pogue concludes with some warnings.

"So Dictate 1.0 is attractive, simple and Mac-like. It is not, however, as good as NaturallySpeaking 9.0 for Windows ($200). It lacks features like audio playback of what you said, a simple “add word” command, legal and medical versions, and non-English language kits."

"It also lacks voice correction."

"When NatSpeak makes an error, you just say “Correct ‘ax a moron’ ” (or whatever it typed); and choose from a list of alternate transcriptions. The program not only corrects the error in your document, but also learns from its mistake. Over time, the accuracy edges ever closer to 100 percent."

"In Dictate 1.0, however, you have to fix transcription errors by hand. The company intends to add voice correction in a 1.1 update; in the meantime, though, your accuracy won’t improve."

The late beta version I tested has some bugs. The company intends to get these fixed by the 1.0 version’s mid-February release.

Bugs in beta-quality shipping version 1.0 fully fixed?
Pogue notes "some bugs" in his kind review. Early buyers had less fortunate experiences visited upon them by the incredibly buggy beta software shipped them as "1.0". See their cries of anguish by searching online, and check whether the problems are fixed in the 1.0.1 release. Try, searching e.g., "macspeech dictate ~crash" on web, in news & blogs, etc. If enough hits, add 1.0.1 to search. Collect results for inclusion in main article.

This on 1.0.1


 * MacSpeech Dictate 1.0.1 from MacSpeech brings to the new speech recognition system better documentation, easier license key handling, and enhanced AppleScript support, including the capability to create commands. Version 1.0.1 also includes a number of bug fixes. Registered users can get it by using the Check for Updates feature within the program. ($199 new (or more, depending on headset), free upgrade)

http://db.tidbits.com/article/9542

Pricy cost/benefit ratio
At $199 (more with good headset), delivers far less, but costs more than $190 Windows version. (And does that $190 Windows price already include a good headset?)

Dragon NaturallySpeaking 11, released August 2010, has correction features that, when MacSpeech has them, will make MacSpeech a much better value
The Dragon NaturallySpeaking 11 correction function at last allows the user to train the program on vocabulary by selecting incorrectly-recognized words and correcting them by speaking or typing. If the selected words are not corrected, but rewritten into something completely different, Dragon NaturallySpeaking 11 recognizes that the change is an edit rather than a recognition correction, and does not change the speech model.

Since Nuance now owns both Dragon NaturallySpeaking and MacSpeech, and both are based on the same voice algorithms, there is reason to hope that this correction-done-right feature will be incorporated into the next version of MacSpeech. When it is, user-friendly speech recognition will at last be available without a lengthy and frustrating process of training the program on whatever words or phrases it does not initially recognize.

Difficulties with corrections & training in MacSpeech Dictate.
7.     mwra      Member

Although you can train one word in the manner you describe, I've found it has no effect. Even after you have trained a word correctly, Dictate then doesn't pick up the 'correction' in it's Recognition window. It looks like a back-of-house bug where the recognition selections overlook new/retrained vocabulary items in favour of the default vocabulary. At least, that's the only logical way I can explain the errors I'm seeing. Posted 1 year ago #

1.SeanW Member

Often, I find that MacSpeech Dictate dictation errors are small. For example, one word out of an entire sentence is in the incorrect tense. Rather than using the train commands, I've done it more effective and efficient to simply say "select the word and then repeating the word, usually dictate gets the correct version on the first try -- and seems to learn from this. Only use the training commands when correcting a whole sentence or phrase, or when adding a word to vocabulary. Posted 1 year ago #

2. Charlie Member

This technique is certainly efficient, however, Dictate will not learn from changes made in this way. In order for it to learn, the correction has to be done in the Training window. Charlie Posted 1 year ago #

3. DoctorK     Member

It's not working for me in the "Training" window. Hours of training on one word - my name!

Posted below in another thread:

I've now spent hours eagerly hoping that the "train the word ____" command would easily and reliably work. Sadly, and even more frustratingly, it doesn't! All I'm trying to do is get it to recognize my name! I say "Kernan" (it's pronounced Kerr Nan) and up comes Conan; Kern and; and multiple others. No matter what technique I've used, nothing works. I keep "training" the word "Kernan" and when the window brings up "Conan", I type in "Kernan" and select it.

I then say my name again, and up comes "Conan". This is going to make me barbaric towards my otherwise grand and brand new iMac.

I have used every suggestion and technique offered in the help manual (very slim guidance) and this forum.

I keep thinking I must be doing something wrong or missing something.

MacSpeech seems to have such promise and gets most of what I say pretty decently. Given that I'm preparing to dictate hundreds of pages of book draft material, I sure would like to get this command down.

Suggestions?Posted 1 year ago #

4.     Charlie      Member

Hey, DoctorK, see my post responding to your duplicate posting on this subject in the Troubleshooting section. Charlie     Posted 1 year ago #

5.     fairgo4all      Member As a longtime Dragon Dictate User I have just come to the Mac world and delighted with the Mac...got excited when I read and thus bought the MacSpeech. Initial recognition terrific and then discovered simple misrecognition or non-recognition errors couldn't be corrected on the fly

I mean it has all been done for ages in the Dragon parent - simply say - "correct that" (meaning the last utterance)and you can immediately correct it by either choosing the correct alternative or training the correct word on the spot.

To have to go through the WORD Recognition process is to seriously interrupt the thought process-- hate to say it but this knocks MacSpeech out of the ring for me - and I will have to use Dragon 10 or even 9 on the notebook pc for creative dictation.

I will experiment some more but notice another user who has spent more time on it (and has experienced the flow and faster/more practical method long developed and available in Dragon) has reached the same conclusion.

If someone who reads this has any influence with the people who build and offer this program please just tell them that this (inability to easily correct errors as you go) is a serious defect in their offering and should be remedied forthwith. I think it gets close to being a dud for this critical reason. I certainly could not recommend it to anyone until this defect has been fixed. Posted 1 year ago #

6.     fairgo4all      Member Training one word - like your name is easily done in the panel where you add words and then train them - although I found I had to re-type the word for training...even though it had been added??

My complaint is the error correction function should be in the flow of dictation - as it is in Dragon Dictate - where all of this came from anyway - or so the literature says. Posted 1 year ago #

7.     mwra      Member Although you can train one word in the manner you describe, I've found it has no effect. Even after you have trained a word correctly, Dictate then doesn't pick up the 'correction' in it's Recognition window. It looks like a back-of-house bug where the recognition selections overlook new/retrained vocabulary items in favour of the default vocabulary. At least, that's the only logical way I can explain the errors I'm seeing. Posted 1 year ago # —Preceding unsigned comment added by 66.167.61.70 (talk) 15:21, 20 September 2010 (UTC)

This entry should state whether MacSpeech Dictate bugs are fixed in "Dragon Dictate for Mac"
MacSpeech Dictate has a bug, encompassed in the "golden rule" workaround requiring users to use no "hands and voice" in the same session (can't just type in a correction for the words it misrecognized and have the program recognize your correction and update its voice model). The result has been train-of-thought shattering hoops to jump through every time. Errors could also not be ignored, because the program would either learn nothing or learn wrong. This and other user interface bugs have been fixed in the Windows version of the Dragon product, but not the Mac version.

MacSpeech also has a cache handling workaround, a cache issue "solved" by a most inconvenient user workaround called the "Golden Cache", so that it requires a steep learning curve to learn how to manage/program the workaround, and attention after learning.

Finally, corrected words don't remain corrected, and laboriously "trained" words won't stay trained.

For these reasons, many buyers have been bitterly disappointed by the failure of MacSpeech Dictate to match the ease promised in the hype. Most put the programs back on the shelf and wrote off their losses.

This section should answer the question of whether the September 20 "Dragon Dictate for Mac" or "Dragon Dictate 2.0" has fixed the serious user interface problems of MacSpeech Dictate.

Hopefully, a "Dragon Dictate for Mac" manual or other detailed description will be made available soon. The failure to release one, and pointing in promotional materials to MacSpeech Dictate's manual, which sets out the unsatisfactory workarounds, is troubling. Hopefully, the lack of a manual does not indicate that the MacSpeech bugs are not fixed. —Preceding unsigned comment added by 68.165.11.115 (talk) 13:29, 21 September 2010 (UTC)

External links modified
Hello fellow Wikipedians,

I have just modified one external link on MacSpeech Dictate. Please take a moment to review my edit. If you have any questions, or need the bot to ignore the links, or the page altogether, please visit this simple FaQ for additional information. I made the following changes:
 * Added archive https://web.archive.org/web/20120324182808/http://www.speech.kth.se/~rolf/gslt_papers/MarkusForsberg.pdf to http://speech.kth.se/~rolf/gslt_papers/MarkusForsberg.pdf

When you have finished reviewing my changes, you may follow the instructions on the template below to fix any issues with the URLs.

Cheers.— InternetArchiveBot  (Report bug) 10:24, 29 May 2017 (UTC)

External links modified
Hello fellow Wikipedians,

I have just modified 5 external links on MacSpeech Dictate. Please take a moment to review my edit. If you have any questions, or need the bot to ignore the links, or the page altogether, please visit this simple FaQ for additional information. I made the following changes:
 * Added archive https://web.archive.org/web/20110101105208/http://www.macspeech.com/article_info.php?articles_id=324 to http://www.macspeech.com/article_info.php?articles_id=324
 * Added archive https://web.archive.org/web/20110101105105/http://www.macspeech.com/article_info.php?articles_id=328 to http://www.macspeech.com/article_info.php?articles_id=328
 * Added archive https://web.archive.org/web/20100108130736/http://www.macspeech.com/article_info.php?articles_id=335 to http://www.macspeech.com/article_info.php?articles_id=335
 * Added archive https://web.archive.org/web/20100124070231/http://www.macspeech.com/article_info.php?articles_id=334 to http://www.macspeech.com/article_info.php?articles_id=334
 * Added archive https://web.archive.org/web/20090215202910/http://www.macspeech.com/articles.php?tPath=7 to http://www.macspeech.com/articles.php?tPath=7

When you have finished reviewing my changes, you may follow the instructions on the template below to fix any issues with the URLs.

Cheers.— InternetArchiveBot  (Report bug) 07:18, 11 January 2018 (UTC)