0

I' m trying to create a complex grammar for Microsoft.speech in c#. I would like to improve accuracy and though, I would like to avoid the DictationGrammar() class. Instead, I would like to create a Grammar() object that will include the entire English vocabulary. Adding all words in the vocabulary as Choices() , the recognizer will not recognize sentences. I am finding problems to create sentences that the engine will recognize. any idea to solve it?

EDIT:

The idea is to be able to communicate with my app as more fluent as possible. The DictationGrammar() class makes a lot of mistakes during the recognition. Probably a custom grammar could decrease the error probability. any ideas?

bre_dev
  • 562
  • 3
  • 21
  • What you are trying to do does not fit the whole idea of the speech recognizer implementation. For large vocabulary you need to use DictationGrammar since it is specificaly designed for large vocabulary. Just adding multiple choices is not going to work because word list information is not enough to provide a good recognition accuracy and speed. You need to describe better the application you want to implement to get an advice on the best algorithm to use. – Nikolay Shmyrev Dec 22 '13 at 09:29
  • I updated the question, even training the speech engine in Windows, the recognizer still makes mistakes. – bre_dev Dec 23 '13 at 21:16
  • You can adapt it to get better accuracy. Most likely the issue is not in grammar. Also you might want to find a way to record more clean speech (without noise and reverberation). Grammar is rarely a cause of accuracy issues. – Nikolay Shmyrev Dec 24 '13 at 09:44
  • What you mean with: "adapt it?" , Adapt what? – bre_dev Dec 25 '13 at 18:56
  • Adapt the recognizer, see for details http://stackoverflow.com/questions/299799/acoustic-training-using-sapi-5-3-speech-api/1522908#1522908 – Nikolay Shmyrev Dec 25 '13 at 21:24
  • I am still trying to increase recognition precision. Turned out that the recognizer adaptation methods are not exposed in c#. What if I train the recognizer using Windows Speech training from control panel? Will the dictation precision be increased? – bre_dev Mar 20 '14 at 22:30

0 Answers0