Re: [Kaldi-users] Reg: Word level recognition using Timit Database

[Daniel P. <dp...@gm...>]

My guess is that you might have a lexicon that does not cover the majority
of whatever words you have in the transcripts (e.g. in data/train/text).
In TIMIT I think the supplied transcription is at at a phone level; if you
use a word level lexicon with that it won't work, and you'll get a lot of
warnings about OOVs in exp/mono/log/compile_graphs.*.log
Dan


On Sun, Jan 18, 2015 at 5:50 AM, Naresh kumar <ell...@gm...>
wrote:

> Dear all,
> I am building word level ASR system using timit database. Currently I am
> using new version of kaldi which is s5. I have reduced the phone mapping to
> 39.  I used 3980 utterances to train language and acoustic models.
>
> I have computed mfcc features for all training files. Monophone training
> was done. The decoding has been done for all the test sentences. After
> decoding I found that all the estimated sentences are very bad (%WER is
> 123.2).
>
> My lexicon looks like this:
>
>
>
>
> *abbreviate      ah b r iy v iy ey tabides  ah b ay d zability ah b ih l
> ah t iy*
> While monophone training I found:
>
>
>
>
>
>
>
> *WARNING (gmm-acc-stats-ali:main():gmm-acc-stats-ali.cc:79) No alignment
> for utterance fpjf0_sx352WARNING
> (gmm-acc-stats-ali:main():gmm-acc-stats-ali.cc:79) No alignment for
> utterance msms0_si1433WARNING
> (gmm-acc-stats-ali:main():gmm-acc-stats-ali.cc:79) No alignment for
> utterance mwad0_si1062WARNING
> (gmm-acc-stats-ali:main():gmm-acc-stats-ali.cc:79) No alignment for
> utterance msms0_si1433WARNING
> (gmm-acc-stats-ali:main():gmm-acc-stats-ali.cc:79) No alignment for
> utterance mwad0_si1062WARNING
> (gmm-acc-stats-ali:main():gmm-acc-stats-ali.cc:79) No alignment for
> utterance mwad0_si1062*
>
>
>
> *1118 warnings in exp/mono/log/acc.*.*.log4680 warnings in
> exp/mono/log/update.*.log43824 warnings in exp/mono/log/align.*.*.log*
> Done
> fsttablecompose data/lang_test_bg/L_disambig.fst data/lang_test_bg/G.fst
> fstminimizeencoded
> fstdeterminizestar --use-log=true
> fstisstochastic data/lang_test_bg/tmp/LG.fst
> 0.693359 -0.0765151
> [info]: LG not stochastic.
> fstcomposecontext --context-size=1 --central-position=0
> --read-disambig-syms=data/lang_test_bg/phones/disambig.int
> --write-disambig-syms=data/lang_test_bg/tmp/disambig_ilabels_1_0.int
> data/lang_test_bg/tmp/ilabels_1_0
> fstisstochastic data/lang_test_bg/tmp/CLG_1_0.fst
> 0.693359 -0.0765151
> [info]: CLG not stochastic.
> make-h-transducer --disambig-syms-out=exp/mono/graph/disambig_tid.int
> --transition-scale=1.0 data/lang_test_bg/tmp/ilabels_1_0 exp/mono/tree
> exp/mono/final.mdl
> fsttablecompose exp/mono/graph/Ha.fst data/lang_test_bg/tmp/CLG_1_0.fst
> fstdeterminizestar --use-log=true
> fstminimizeencoded
> fstrmsymbols exp/mono/graph/disambig_tid.int
> fstrmepslocal
> fstisstochastic exp/mono/graph/HCLGa.fst
> 0.693359 -0.0761903
> HCLGa is not stochastic
> add-self-loops --self-loop-scale=0.1 --reorder=true exp/mono/final.mdl
> steps/decode.sh --nj 5 --cmd run.pl exp/mono/graph data/dev
> exp/mono/decode_dev
> decode.sh: feature type is delta
> steps/decode.sh --nj 5 --cmd run.pl exp/mono/graph data/test
> exp/mono/decode_test
> decode.sh: feature type is delta
>
>
> Please let me know how to solve this issues of all the warnings and to
> reduce the  WER.
>
> --
>
> Regards
> Naresh Kumar
>
>
> ------------------------------------------------------------------------------
> New Year. New Location. New Benefits. New Data Center in Ashburn, VA.
> GigeNET is offering a free month of service with a new server in Ashburn.
> Choose from 2 high performing configs, both with 100TB of bandwidth.
> Higher redundancy.Lower latency.Increased capacity.Completely compliant.
> http://p.sf.net/sfu/gigenet
> _______________________________________________
> Kaldi-users mailing list
> Kal...@li...
> https://lists.sourceforge.net/lists/listinfo/kaldi-users
>
>