Using Dragon with two languages
by Volker Weber
When I first installed Dragon, I did not know how to install two languages. I have the German version, so choosing German as the language was the obvious choice. Later I found out, that you can create a new profile, and then select another language. By switching the profile, you can now choose the language for dictation. With the language you're using, not only the dictation language changes, but also the commands for the software. You are basically learning two software packages.
When you're using two distinct setups, you also learn about the differences. With the German version my dictation is way more accurate. The reason is simple: I can speak German much clearer than I can speak English. When you set up English as the language, you have to choose between different regions where English is spoken, and you also have to select an accent. I chose American, because that is where I speak most of my English. Not being a native speaker of course means that I don't have a clear accent. While the dictation results are better in German, the commands are not. I blame the sloppy translation between the original English version of the software and the German one.
Comments
I have often wondered about using Dragon to dictate while coding. I can see/imagine the obvious benefits when it comes to writing. Has anyone ever tried Dragon while coding?
Do you mean you would like to write your code with dictation? Are you a COBOL programmer? Or would you like to multitask with coding and dictation at the same time?
This Dragon always codes. Couldn't guarantee the outcome of the final product though. :)
As an aside I tried Dragon Dictate many, many years ago. I found it awkward to use. Not tried it recently so hopefully it has improved somewhat. There is the old classic test of whether it can understand the difference between "It was a grey day" and "It was a Grade A". I suppose it's all about the pronunciation. Anyhow I ended up dropping it because there were just too many errors and it took more time to correct said errors than it would to actually type it in the first place.
Yes - I was thinking that being able to dictate code (Swift - iOS) in Xcode could be very effective. I am a visual person and tend to plan out what I'm doing on paper using mind maps. If I could focus on the map and then "speak" the code with out having to look at the monitor that would be great. Or am I some years ahead in my thinking?
Yes, you are. Speech recognition has advanced from phonetics to speech models. The software understands sentences. It will often correct itself once you finish a sentence. So it kind of knows what you are going to say. There are certain specialized language models, like those for doctors or lawyers, who have a very controlled vocabulary. The developer would have a very different speech model for his programming language. I don't think that there is a market yet to develop this kind of model.
If you look at the screenshot, you will see that it's currently in dictation mode. There are other modes like spelling mode, where the software expects a different vocabulary. In this mode you only speak numbers, punctuation marks, or single characters. Imagine that you would need another mode just for programming.
@Volker - Thank you. What you say makes perfect sense. It is a shame, there is at least a market of 1 for that product.
Am I the only one imagining now people making interesting noises, like this guy here - who used speech recognition to write code?
https://www.youtube.com/watch?v=8SkdfdXWYaI
No idea how Dragon works, but in general this is a machine learning exercise and about probabilities, meaning in a document certain words are likely to follow other certain words. That's what spelling software understands and it is what makes recognition easier, because you can use those probabilities to tune the recognised results.
When narrowing down the context the recognition gets better, for example learning from Volker's mails and texts how he uses language.
So taking up what Ian says this should be even better for coding. Here the expected syntax and structure is extremely rigid. If speech recognition software learns this context, it should be easier to understand code than arbitrary text, I think.
Navigation in code by voice seems very slow to me, but for Ian's case, translating pre-structured thoughts into code, that is not really all that important anyway.
@Daniel - that's a great find. What a great story and solution. Of course it also highlights that really it's never going to happen for something like Xcode.
is is best to buy dragon for mac in their store (175,- usd)? are all languages (german) included in that version too?
No, you need Dragon Deutsch.
