A few days ago, I posted a review of Dragon Dictate for Mac. Tellingly, it was entitled “utterly frustrating” and I illustrated my frustration with a video of the product in action (included below). Today, Nuance (maker of Dragon Dictate) responded with a detailed comment. Since their comment was long and thoughtful, I decided to post it, and my response, as a new post.
My Feelings About Dragon
I really wanted to like this product. Seriously. I’d heard so many great things about Dragon Naturally Speaking for Windows but could never afford it. Once I switched to Mac, I was concerned about the MacSpeech until Nuance bought the company and updated it as Dragon Dictate 2.0. So I decided to take the plunge when Amazon offered it as a “Gold Box” deal.
Since then, I have been completely frustrated with the product. It does a great job recognizing speech, but the interface is so limited and clunky it’s just about impossible to use. There’s nothing wrong with the underlying engine, just the packaging of it into a Mac application. If my basic concerns were addressed, I bet it would become an excellent product.
The Dragon Responds
Erica Hill from Dragon corporate communications left a comment in my post, hoping “to hear back from you, either here or directly via email.” So here’s your response, Erica!
First, thank you for responding at all. All too often, companies don’t have this kind of proactive, honest response to feedback, especially when it’s negative. But you’re doing the right thing by responding like this!
Thanks for your thoughtful review of Dictate. We sincerely appreciate all feedback — good or bad, as it only helps us to improve Dictate.
I’m sorry that you haven’t had a good experience. We don’t hear this too often, but I’m glad that you’ve let us know, as it doesn’t matter if it’s one person or 100 — the experience should be enjoyable and productive for everyone. You mentioned wanting a refund — we will certainly oblige — just drop me a line at erica DOT hill AT nuance DOT com and I’ll get the process going for you.
I actually don’t want a refund. I want a functioning product. How about an upgrade to the new, improved, functional version of Dictate you so must be feverishly working on in the back room? Barring that, I suppose a refund might be the only recourse if the product will remain as-is for the foreseeable future.
Bottom line, your points are taken, and please know that we are committed to improving Dictate with every new version and update, so your feedback helps. With regard to your suggestions for improvements, I checked in with the product team and have a few thoughts to share on some specific points, and a question or two.
My suggestion: Add a simple, streamlined dictated command to add a word to the dictionary — “add that” should allow you to spell and train in one go
Use “Access or Show Vocabulary Editor Window” to open the Vocabulary Editor window. But you are correct, there is no voice-only way to add custom vocabulary words. But once you add a custom word like “Fibre Channel” and train it, you should see that word or phrase recognized in your dictation.
Ok, so it’s possible to kick off the complex process of adding a word with a dictated command. Thank you! Now what about the rest of the process? This would be a very useful feature. As for recognition, as you can see, it sometimes recognizes added words correctly and sometimes not. It’s all very frustrating. For example, even though I’ve added and trained “SATA” (pronounced as “SAT-uh”) it never, ever recognizes it correctly. Not once.
Use “Train Vocabulary From Selection” after selecting a word or words using the “Select…” command.
I’m not clear on what this does. I knew about this command, but it didn’t appear to do anything at all. When I try this, I invariably get a window that says “all the words in the text supplied are already in this vocabulary.” And it does not appear that this improves recognition accuracy or adds words to the vocabulary.
For example, I recently wrote an article about the Micron P320h SSD. I did not add that long, clunky product name to the Dictate vocabulary, so every time it appears I either have to say the ungainly phrase, “cap P no space 320 lowercase h” or type it in manually. Yet selecting that name and saying “train vocabulary from selection” pops up a window claiming that all the words are in the vocabulary. They are not, and are not recognized correctly. So I don’t understand what the point is.
Use “Vocabulary Training” to scan existing documents for unique vocabulary words.
This seems to work to add words to the vocabulary, but it does not appear to train the pronunciation of them. And it seems to skip words like “P320h”. I will certainly begin using this function in the future, but I don’t hold out much hope that it will improve recognition accuracy for technical terms.
My suggestion: Fix the obnoxious behavior outside the Dictate “notepad” window — or add an “insert only” setting to not “guess” where you are in a document and overwrite things
“Universal Tracking” is something we are working on. Currently we do recommend that folks dictate within TextEdit for the best experience.
My suggestion is simpler than universal tracking as defined here. Rather than trying to make the application smarter, why not make it dumber? Don’t automatically insert spaces or retrace characters in other applications than TextEdit. Just insert what is dictated and let the user handle cursor movements. This would be much more intuitive than the bizarre behavior exhibited by Dragon Dictate currently.
Where exactly do you recommend dictating in TextEdit? The online marketing materials claim “watch your words appear on screen in your favorite Mac applications” and “Dragon Dictate for Mac works with your favorite Apple applications, including Mail, iChat, iCal, TextEdit, Pages, Safari and many more.” I see no mention of Dragon Dictate’s reliance on TextEdit on your website or on the product box. Rather, promotional materials indicate that the product works in any application.
One commenter on my blog suggested using the “cache selection” command to improve Dictate’s behavior in other applications. I will try this in the future as well.
My suggestion: Fix the weird “add extra characters after the cursor” bug demonstrated in my video
This will be addressed in a future release.
That’s good to know. Although you claim that most users are satisfied with the program, this bug is widely documented on the Internet for years, and I have received condolences from other users who say this is the main reason they stopped using the program. Fixing this problem should be priority one for Nuance, since it affects every user in every application.
My suggestion: Allow me to fix one or two words, not a whole phrase, and allow me to fix it outside the Recognition window’s (limited) choices
These are great suggestions and the correction process is something that we are always striving to improve.
The recognition window issue is a major pain. Often, it simply refuses to offer an appropriate choice of words, even after multiple attempts at dictating the word desired. This happens with known words as well as those not in the vocabulary editor. It would be tremendously useful to be able to say “spell that word” or something similar rather than relying on the choices that the recognition window shows.
My suggestion: Add a “no abbreviations” setting to the Mac version so it stops correcting “minutes” to “min.”
Finer control of engine settings for AutoFormatting will be addressed in a future release.
This is very good news indeed, since it is majorly frustrating that Dictate will not allow me ever to dictate the word “minutes” after a number. It always changes my wording to “15 min.”
At the very least, the recognition window should offer the spelled-out version of any abbreviation as a optional correction. I would think this would be a very simple feature to add. But at present, recognition only offers one choice: The abbreviated form that has already been recognized.
Weighted Recognition Options
My suggestion: Allow me to say “always recognize it this way” for words like “Fibre Channel” and “Stephen’s stance” (not “fiber channel” and “Steven’s stance”)
With addition of custom vocabulary words, items like “Fibre Channel” will appear in dictation. Sample dictation:
1. Let’s review the latest in Fibre Channel designs for high-performance storage area networks. I would like to know Stephen’s stance on this issue.
I have added these the vocabulary words to Dictate, but it still recognizes them incorrectly the majority of the time. As illustrated in the video, “Fibre Channel” is often recognized correctly but sometimes not, even though the alternative “fibre channel” would never appear in a document I was dictating.
Again, my suggestion is to allow the user to say “never recognize that” or something similar.
My suggestion: Improve recognition of common commands and terms (“explanation point”? Seriously? When has anyone ever dictated that?)
Would like to know if you were using the mic that came with Dictate or a different audio input source. This type of misrecognition could be due to poor audio quality. Can you let me know?
I am using the headset mic and USB adapter that came with Dictate. However, when using this mike in other applications like Skype, I have heard that sound quality is poor. Perhaps there is something wrong with the mic hardware?
More Issues with Dictate for Mac
Although I did not comment on these issues at length, since we’re talking about improvements to the application, let me take a few moments to outline some more issues that I am experiencing.
- High CPU utilization: Dictate will often use quite a bit of my system’s CPU, pinning one of the four cores near 100%. This seems excessive, especially since it does not always require this much juice. I suspect a bug.
- Delayed recognition: After a while of use, Dictate begins slowing down. Just now, it was taking 30 seconds or more for my words to appear in the floating window and another 20 seconds or so to appear in the document. I have to restart Dictate or just return to the keyboard. This is often a prelude to the third issue…
- Crashing: Dictate has proven unstable in regular usage, often crashing after extended usage. I usually allow the application to send these reports to Nuance, so you should have this information already. When it crashes, Dictate often loses a few minutes of work in the TextEdit window.
That Dragon chose to respond to my issues publicly, and am willing to work with you to resolve these problems. But it seems that there is no solution at present to most of my concerns, and the additional issues of load, slowdown, and crashing do not inspire confidence. I still cannot recommend this application, and find myself increasingly frustrated with it.
What I would like from Dragon mainly is a roadmap showing definite improvements in the near future. Resolve the random character issue (which has been known for years), fix the slow down and crashing issue, and perhaps add some of the easier feature requests and streamline the process of adding vocabulary. All of these should be doable for a company like Nuance, especially since the product retails for so much money.
When can we expect these fixes? Are you working on them? We customers would love to know.