No alignment result for librispeech example dataset

[manazhao]

Binary: 1.0.1 Linux.

Command: I was following the librispeech example.

bin/mfa_align -v -d -c \
 /data/asr/librispeech/corpus/19 \
 /data/asr/librispeech/lexicon.txt \
 english \
 /tmp/alignment

No runtime errors were seen on the output and the output folder /tmp/alignment only contains two files: oovs_found.txt utterance_oovs.txt. I checked the log ( ~/Documents/MFA/19/logging/corpus.log) and saw the following,

Setting up corpus information...
Number of speakers in corpus: 1, average number of utterances per speaker: 111.0
Setting up training data...
The following utterances were ignored due to lack of features: 19-198-0000, 19-198-0001, 19-198-0002, 19-198-0003, 19-198-0004, 19-198-0005, 19-198-0006, 19-198-0007, 19-198-0008, 19-198-0009, 19-198-0010, 19-198-0011, 19-198-0012, 19-198-0013, 19-198-0014, 19-198-0015, 19-198-0016, 19-198-0017, 19-198-0018, 19-198-0019, 19-198-0020, 19-198-0021, 19-198-0022, 19-198-0023, 19-198-0024, 19-198-0025, 19-198-0026, 19-198-0027, 19-198-0028, 19-198-0029, 19-198-0030, 19-198-0031, 19-198-0032, 19-198-0033, 19-198-0034, 19-198-0035, 19-198-0036, 19-198-0037, 19-227-0000, 19-227-0001, 19-227-0002, 19-227-0003, 19-227-0004, 19-227-0005, 19-227-0006, 19-227-0007, 19-227-0008, 19-227-0009, 19-227-0010, 19-227-0011, 19-227-0012, 19-227-0013, 19-227-0014, 19-227-0015, 19-227-0016, 19-227-0017, 19-227-0018, 19-227-0019, 19-227-0020, 19-227-0021, 19-227-0022, 19-227-0023, 19-227-0024, 19-227-0025, 19-227-0026, 19-227-0027, 19-227-0028, 19-227-0029, 19-227-0030, 19-227-0031, 19-227-0032, 19-227-0033, 19-227-0034, 19-227-0035, 19-227-0036, 19-227-0037, 19-227-0038, 19-227-0039, 19-227-0040, 19-227-0041, 19-227-0042, 19-227-0043, 19-227-0044, 19-227-0045, 19-227-0046, 19-227-0047, 19-227-0048, 19-227-0049, 19-227-0050, 19-227-0051, 19-227-0052, 19-227-0053, 19-227-0054, 19-227-0055, 19-227-0056, 19-227-0057, 19-227-0058, 19-227-0059, 19-227-0060, 19-227-0061, 19-227-0062, 19-227-0063, 19-227-0064, 19-227-0065, 19-227-0066, 19-227-0067, 19-227-0068, 19-227-0069, 19-227-0070, 19-227-0071, 19-227-0072.  See relevant logs for more information
Number of speakers in corpus: 1, average number of utterances per speaker: 0.0

It seems to me mfcc features were not properly created for the audio files, though the audio files look normal to me.

Any suggestion for addressing this issue will be greatly appreciated!

[elwinmt]

Same problem here

[YuhaoT]

Same issue here. I use the following command which copy from the tutorial except that I changed the paths.
bin/mfa_align ~/montreal-forced-aligner/docs/English\ Files/Librispeech ~/montreal-forced-aligner/dictionary/librispeech-lexicon.txt english ~Documents/aligned_librispeech

After run in terminal, it generate two folders /MFA and /aligned_librispeech. The output directory aligned_librispeech has nothing. Can anyone give me a further instruction about how to use the aligner? Thanks in advance!

[Liujingxiu23]

Same problem here, can someone help?

log show align successfully, but nothing in the output dir.

INFO - Setting up corpus information... lINFO - Number of speakers in corpus: 3, average number of utterances per speaker: 122.33333333333333 INFO - Parsing dictionary without pronunciation probabilities without silence probabilities INFO - Creating dictionary information... INFO - Setting up training data... Generating base features (mfcc)... Calculating CMVN... INFO - Done with setup! INFO - Performing first-pass alignment... INFO - Calculating fMLLR for speaker adaptation... INFO - Performing second-pass alignment... INFO - All done!

[mmcauliffe]

@Liujingxiu23 what version are you using? The previous issues were with 1.0.1, which should be fixed in the 2.0 alpha. Can you try rerunning it with the --clean flag? Can you attach your align.log from the MFA temp directory?