 |
KnowBrainer Speech Recognition |
 |
 |  |
|
Topic Title: Natink & Dragonfly, DPI 15 troubles Topic Summary: Created On: 09/04/2016 05:08 AM Status: Post and Reply |
|
|
 09/04/2016 05:08 AM
|
|
|
I've upgraded to DPI 15. The good news: Natlink registers, and loads. Grammars are recognized, and the simplest commands (Dictation to Key) work. The problem: I haven't managed to get more complex commands to work. Even the simple ones such as "up 5" to move up 5 times. The following is shown in the log:
Other commands don't write anything to the log, but simply have no effect. Does anyone experience the same, or something else? Any ideas? |
|
|
|
|
|
09/04/2016 10:56 AM
|
|
|
More information:
It seems there are two distinct issues. No1: It seems like NatlinkStatus::getLanguage returns zzz, and reports this error: getLanguage: Warning, no model specification string for key 2 1 found in "Base Acoustic" of inifile: C:\ProgramData\Nuance\NaturallySpeaking15\Users\josh\current\acoustic.ini At first glance, this doesn't make sense, as the acoustic.ini file has this section: [Base Acoustic] 2 1=UK English | BestMatch IV The acoustic.ini file in DPI 14 had this section: [Base Acoustic] 0 1=UK English | BestMatch IV 2 1=UK English | BestMatch IV The 2 1 key seems to exist in both. I've tried manually returning "enx" as the language instead of zzz, but it had no effect. No2: Enabling debug callback, I can see the following error whenever there's a recognition which has no effect: Traceback (most recent call last): File "C:\NatLink\NatLink\MacroSystem\core\natlinkmain.py", line 577, in beginCallback (cbd, checkAll, checkForGrammarChanges) natlink.NatError: A SRERR_VALUEOUTOFRANGE error occurred calling ISRResGraph::GetWordNode from ResObj.cpp 186. I couldn't figure out much more about what this means. Any ideas? |
|
|
|
|
|
09/05/2016 01:01 PM
|
|
|
This, together with inabliity to change the microphone icon color and blank Command Browser reported in other posts, is reason enough to stick with DPI 14, unless and until fixes or workarounds are found. |
|
|
|
|
|
09/05/2016 04:30 PM
|
|
|
Switching back to Ver. 14 sounds a little drastic when you consider that you are giving up the Deep Learning engine and double speed dictation. You can easily eliminate the microphone color problem by using the new auto collapse Modern microphone icon view. Admittedly, we are having some issues with getting Dragon to recognize some of our personal commands because Dragon doesn't like abbreviations such as DS-7000 but we assume this is a bug that will be fixed. ------------------------- Forum Mission Statement |
|
|
|
|
|
09/05/2016 06:15 PM
|
|
|
Not switching back, since I haven't installed DPI 15 yet. Just delaying unless and until fixes or workarounds are found.
|
|
|
|
|
|
09/06/2016 04:01 AM
|
|
|
Changing the acoustic.ini file from utf-8-bom encoding to utf-8 made problem No 1 go away. I wouldn't know how to solve it from the Natlink level though, since GetProfileSectionW is not available in pywin.
Problem 2 persists. It only occurs with multiple-word commands, which makes sense, since we're getting a value out of range error. |
|
|
|
|
|
09/06/2016 04:50 AM
|
|
|
At least for me, Natlink not working is a HUGE deal. Pretty much my entire workflow depends on it. I would switch back to 14 were it not for the very fast response by a natlink dev. |
|
|
|
|
09/06/2016 01:04 PM
|
|
|
Not a bug, it's by design. Enter "DS 7000" as written form, and change the printed form under properties to "DS-7000".
-------------------------
|
|
|
|
|
|
09/06/2016 11:07 PM
|
|
|
Nice catch Rüdiger. We hadn't considered these changes to be deliberate. For example, the only way we could get our DS 7000 Hyperlink command to work was to change its name to DS seven thousand Hyperlink; including applying lowercase to “seven thousand” because it didn't work in Title Case. 
------------------------- Forum Mission Statement |
|
|
|
|
|
09/06/2016 11:08 PM
|
|
|
Lunis, can you please move this to the Third-Party Command Utilities - Vocola, Unimacro, VoicePower, Python... forum where it belongs?
|
|
|
|
|
|
09/06/2016 11:15 PM
|
|
|
can you give some examples of non-recognized utterances? The code looks like the problem might be an inadequately sized word buffer.
|
|
|
|
|
|
09/09/2016 02:38 PM
|
|
|
It's a huge deal for me, too. Did you get it to work? If so, how? |
|
|
|
|
|
09/09/2016 04:31 PM
|
|
|
"Traceback (most recent call last):
File "C:\NatLink\NatLink\MacroSystem\core\natlinkmain.py", line 577, in beginCallback (cbd, checkAll, checkForGrammarChanges) natlink.NatError: A SRERR_VALUEOUTOFRANGE error occurred calling ISRResGraph::GetWordNode from ResObj.cpp 186. " I believe this is the NatLink version of the bug people are seeing using the normal SDK here: https://www.knowbrainer.com/forums/forum/messageview.cfm?catid=4&threadid=22974&enterthread=y |
|
|
|
|
|
09/09/2016 06:07 PM
|
|
|
Yes, I told you so via email.
-------------------------
|
|
|
|
|
|
09/10/2016 02:39 PM
|
|
|
Caligula:
I am unclear on what the precise problem is with DPI 15 and Natlink/Vocola. Are you saying that two or more word commands do not work? And you fixed it my editing the acoustical.ini file? |
|
|
|
|
|
09/13/2016 04:11 AM
|
|
|
No, there were two different problems - one, the acoustic.ini file wasn't parsed by Natlink. Changing the encoding fixed that. Secondly - no multi-word commands work with Natlink, and apparently other SDK applications too. This one isn't solved, and sadly renders DPI15 pretty much useless to me and others. |
|
|
|
|
|
09/13/2016 01:04 PM
|
|
|
Bummer. |
|
|
|
|
|
09/13/2016 08:22 PM
|
|
|
Secondly - no multi-word commands work with Natlink, and apparently other SDK applications too. This one isn't solved, and sadly renders DPI15 pretty much useless to me and others. Would it be possible to combine command names as a temporary workaround? For example, could you get away with changing a command name like Insert Paragraph to InsertParagraph? ------------------------- Forum Mission Statement |
|
|
|
|
|
09/14/2016 01:05 AM
|
|
|
Maybe on some of the very simplest of commands, but command chaining ala Dragonfly (Continuous Command Recognition) isn't that simple. The whole point is to allow the user to dictate commands in the same way you'd dictate text. The power isn't just in combining commands, but also in not having to pause after each one.... Anyhow, to give you an example, I've got ~180 commands in my biggest CCR module -- any of them can be chained with any other up to 16 commands deep -- big matrix In short, definitely not. |
|
|
|
|
|
09/14/2016 01:53 AM
|
|
|
Does anyone know if anyone is working to resolve this issue (I'm assuming this would be someone working on NatLink)? |
|
|
|
|
Usually it's only three to five commands at a time, but still, without the chaining, well, you'd be doing something _entirely_ different.FuseTalk Standard Edition v4.0 - © 1999-2021 FuseTalk™ Inc. All rights reserved.