Aquileo | Recent changes to Project InproTK

Aquileo | How can i stop the GUI

Tue, 19 Mar 2019 15:59:08 -0000

it worked .thanks timo i really appreciate ur help :) .

Aquileo | How can i stop the GUI

Tue, 19 Mar 2019 13:10:15 -0000

please try without the -C switch

Aquileo | How can i stop the GUI

Tue, 19 Mar 2019 12:29:47 -0000

Thanks for your quick response.Still I can't block the GUI from appearing and I have attached a screen shot of the lines I have commented in "frontend.xml" link of the screen shot https://drive.google.com/file/d/1gKpZooPsu-cL-_t1XMsHZowd4C_yZQ-i/view?usp=sharing

Aquileo | How can i stop the GUI

Tue, 19 Mar 2019 11:45:21 -0000

Hi Ahmed, I believe the GUI you're seeing is the smiley that shows speaking state, right? You can block that by removing the references to SpeechStateVisualizer from the frontend config). Let me know if this helps. t.

Aquileo | How can i stop the GUI

Tue, 19 Mar 2019 11:37:28 -0000

Hallo,
is there any option to stop the pop-up GUI i traced the code from :
1. Demo_EN_SimpleRecoFromMicrophone.launch
2. SimpleReco.java
3. RecoCommandLineParser.java i tried several output options -O -L -T but the GUI window still there
4. config-en.xml in the demo
java is not my cup of tea ,i am a python developer and really struggle with this code so can you help me.

Aquileo | Inprotk for Turkish

Thu, 16 Mar 2017 12:48:34 -0000

Hi Timo,
We also tried inprotk by replacing multiple tabs with single, and removing ':' characters, however, we got the same error. WO, we could not make it work.

By the way: is inprotk using sphinx4 engine under the hood?
If yes, why our sphinx4 test env does the recognition so much slower than yours? I mean what makes inprotk process so much faster than sphinx4?

Regards

Aquileo | Inprotk for Turkish

Thu, 16 Mar 2017 12:29:52 -0000

Hi Timo,
Thanks for your reply.
First of all let me tell that our AM/dict/LM combinations works pretty well with Sphinx although the recogntion speed is very slow. The accuracy is about 50, which is OK for us at this stage. Currently we are working on our language and acousting model to improve with a lot of trainings.
At this stage, first of all we want to make our model works on inprotk as well.

I tried changing inpro.sphinx.linguist.dictionary.SyllableAwareFullDictionary to inpro.sphinx.linguist.dictionary.Sphinx, it did not work. It comlpains that it can not find component dictionary.
Then I tried it with inpro.sphinx.linguist.dictionary.SyllableAwarePronouncation as well, it did not work either with the same error message.

In the source code that I pulled from the github, the package inpro.sphinx.linguist.dictionary has only SylabbableAwareFullDictionary and Pronouncation. So, I would expect Pronouncation type dictionary would have worked, but it did not.

The error message stack trace :

Setting up monitors...
class not found !java.lang.ClassNotFoundException: inpro.sphinx.linguist.dictionary.SyllableAwarePronouncation
Exception in thread "main" Property exception component:'ngram' property:'dictionary' - component 'dictionary' is missing
edu.cmu.sphinx.util.props.InternalConfigurationException: component 'dictionary' is missing
at edu.cmu.sphinx.util.props.PropertySheet.getComponent(PropertySheet.java:292)
at edu.cmu.sphinx.linguist.language.ngram.SimpleNGramModel.newProperties(SimpleNGramModel.java:99)
at edu.cmu.sphinx.util.props.PropertySheet.getOwner(PropertySheet.java:508)
at edu.cmu.sphinx.util.props.PropertySheet.getComponent(PropertySheet.java:290)
at edu.cmu.sphinx.linguist.lextree.LexTreeLinguist.newProperties(LexTreeLinguist.java:308)

Best Regards

Aquileo | Inprotk for Turkish

Sat, 11 Mar 2017 16:50:15 -0000

Hi Yilmaz, the LM and dictionary look good. However, the problem could be in the lexicon or the combination with the acoustic model. I notice that your lexicon uses multiple whitespace characters between word and list of phonemes. Could you replace them by one single tab (\t)? I reckon you use a turkish AM as well (which I don't know hence can only speculate about). Is the list of phonemes in your lexicon identical to the list of phonemes used by the acoustic model? (The AM should contain a file that lists all phones if it was made with SphinxTrain.) I see a symbol ":" in the dictionary which might not be a phoneme but instead a marker to indicate lengthening of the preceding phoneme?

Does your AM/dict/LM combination work in plain Sphinx-4?

Another source of problems could be my implementation of the dictionary (inpro.sphinx.linguist.dictionary.SyllableAwareFullDictionary). You could change that to Sphinx' default implementation to exclude it as a source of error.

Keep me posted!

Aquileo | Inprotk for Turkish

Wed, 08 Mar 2017 12:00:50 -0000

And here is my config.xml contents :

&sphinx;
&sphinxlang;
&frontend;
&iuconfig;

</config>

sphinx_tr.xml config file is attached.

Aquileo | Inprotk for Turkish

Wed, 08 Mar 2017 11:58:28 -0000

Hi again,
I am attaching language model files for your reference. tutanakSort.arpa.truncated is the truncated version of the actual language model file. I truncated it to avoid wasting resources.

Regards