Cannot get the textgrid result when the input_folder only has only one pair of data(one wav and corresponding lab file)

[quanzzz123]

Hello, thanks for the great job! Here comes one problem in my case. I just want do forced alignment on only one input .wav file with .lab, however there is no textgrid output in the output folder. And I test some other cases. If I put two different pairs of data( different name and different .wav), I will get the correct result. If I use the same pair of data with different names, I will not get any result. So I go to check the logs. In the "align.0.0.log" comes the error as shown below:

LOG (gmm-boost-silence[5.4.251~1-094d2]:main():gmm-boost-silence.cc:103) Wrote model to - LOG (gmm-align-compiled[5.4.251~1-094d2]:main():gmm-align-compiled.cc:127) 01 WARNING (gmm-align-compiled[5.4.251~1-094d2]:AlignUtteranceWrapper():decoder-wrappers.cc:466) Retrying utterance 01 with beam 40 WARNING (gmm-align-compiled[5.4.251~1-094d2]:AlignUtteranceWrapper():decoder-wrappers.cc:475) Did not successfully decode file 01, len = 361 LOG (gmm-align-compiled[5.4.251~1-094d2]:main():gmm-align-compiled.cc:135) Overall log-likelihood per frame is nan over 0 frames. LOG (gmm-align-compiled[5.4.251~1-094d2]:main():gmm-align-compiled.cc:137) Retried 1 out of 1 utterances. LOG (gmm-align-compiled[5.4.251~1-094d2]:main():gmm-align-compiled.cc:139) Done 0, errors on 1

It seems like when using beam 40 , there is something wrong with retrying utterance. For the two different pair of data case, there is no WARNING information.

So, can someone help me with this? Also I will go to the source code to try to fix it.

[simlmx]

I had the same issue. Making sure all the words are in the dictionary fixed it for me (i.e. no prompt to fix words not in the dictionary and an empty oovs_found.txt file). I think it can tolerate some out-of-vocabulary words but I don't know what the threshold is.

[quanzzz123]

I had the same issue. Making sure all the words are in the dictionary fixed it for me (i.e. no prompt to fix words not in the dictionary and an empty oovs_found.txt file). I think it can tolerate some out-of-vocabulary words but I don't know what the threshold is.

Thanks. Actually the words of my lab file and wav all exist in the dictionary, maybe its the problem of Kaldi align shell scripts? I have no idea of it.

[simlmx]

@quanzzz123 Yeah turns out I encountered the same problem (no output) at times where all my words are in the vocabulary. No idea what is going on either!

[GalaxieT]

@simlmx Same problem. Turns out that I was using some wrong lab file. It seems that things would go well if there weren't so many words out of the dictionary.