Open dpny518 opened 5 years ago
Hi @yondu22,thanks for pointing out that the project could have better decoders. Given that the spare time I have, probably I will start to implement the new feature from next month.
@rayworks Thanks for this repo. I found a timer functionality in the mainactivity class. Besides that what does this project do? Is the goal to run kaldi on device (offline) like https://keenresearch.com/ ??
@nanaghartey, the original purpose for this project is to have a PoC using kaldi on Android platform in offline mode.
Before this, we've used CMU PocketSphinx for a long time already, but recently we just found it doesn't work as expected for some cases.
@rayworks alright but currently it seems there is no decoding functionality and as such speech recognition can't be performed like pocketsphinx right?
No, not on the JNI level.
Alright, Do you plan to add those features on the JNI level to it?
On Thursday, July 18, 2019, 1:06:33 PM GMT, rayworks <notifications@github.com> wrote:
No, not on the JNI level.
— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub, or mute the thread.
Speaking of decoders, have you tried with this one? I will continue working on current project to fulfill a meaningful demo instead of an intermediate one. But I can't guarantee the completion time.
@yondu22 Thanks for the link. It builds and runs successfully however, the accuracy is poor. I keep getting wrong results. Did you face same issue?
Could you try to add JNI wrappers like this https://github.com/ahmetaa/kaldi-jni/blob/master/cpp/jni/kaldi-jni.cc
for this code https://github.com/alphacep/kaldi-websocket-python/blob/master/kaldi_recognizer.cc
I think this will allow android decoding