Closed kimkwangho82 closed 4 years ago
Thank you for your effort and achievement. I have studied Kaldi (Speech-Recognition Tool) for years.
If you read the below Kaldi code, the word's start_time and end_time is not same between below code and 'decoder.hpp' code.
Kaldi code : https://github.com/kaldi-asr/kaldi/blob/master/src/online2bin/online2-tcp-nnet3-decode-faster.cc#L324
In 'decoder.hpp' code, 'frame_subsampling' is missed. could you check the decoder's word's start_time and end_time.
Thank you.
@pskrunner14 i am assuming you might have better view on this?
Hi @kimkwangho82 thanks for pointing out the bug! It has been fixed in v0.3.0
Thank you for your effort and achievement. I have studied Kaldi (Speech-Recognition Tool) for years.
If you read the below Kaldi code, the word's start_time and end_time is not same between below code and 'decoder.hpp' code.
Kaldi code : https://github.com/kaldi-asr/kaldi/blob/master/src/online2bin/online2-tcp-nnet3-decode-faster.cc#L324
In 'decoder.hpp' code, 'frame_subsampling' is missed. could you check the decoder's word's start_time and end_time.
Thank you.