Added comments to RM data_prep/run.sh script; changed some filenames in mkgraph.sh

git-svn-id: https://svn.code.sf.net/p/kaldi/code/trunk@90 5e6a8d80-dfce-4ca6-a32a-6e07a63d50c8
This commit is contained in:
Dan Povey 2011-06-17 19:44:56 +00:00
Родитель 2399143086
Коммит 2c6b83925f
3 изменённых файлов: 15 добавлений и 5 удалений

Просмотреть файл

@ -19,6 +19,15 @@
# The script's argument is a directory which has three subdirectories:
# rm1_audio1 rm1_audio2 rm2_audio
# Note: when creating your own data preparation scripts, it's a good idea
# to make sure that the speaker id (if present) is a prefix of the utterance
# id, that the output scp file is sorted on utterance id, and that the
# transcription file is exactly the same length as the scp file and is also
# sorted on utterance id (missing transcriptions should be removed from the
# scp file using e.g. ../scripts/filter_scp.pl)
# You get get some guidance how to deal with channels and segments (not
# an issue in RM) from ../scripts/make_mfcc_train_segs.sh.
if [ $# != 1 ]; then
echo "Usage: ./run.sh /path/to/RM"
exit 1;

Просмотреть файл

@ -78,12 +78,12 @@ fsttablecompose $dir/ilabel_map.fst $dir/CLG.fst | fstdeterminizestar --use-log
cat $dir/CLG2.fst | fstisstochastic || echo "warning: CLG2 is not stochastic."
make-h-transducer --disambig-syms-out=$dir/disambig_tstate.list \
make-h-transducer --disambig-syms-out=$dir/disambig_tid.list \
--transition-scale=$tscale $dir/ilabels.remapped $tree $model > $dir/Ha.fst
fsttablecompose $dir/Ha.fst $dir/CLG2.fst | fstdeterminizestar --use-log=true \
| fstrmsymbols $dir/disambig_tstate.list | fstrmepslocal | fstminimizeencoded > $dir/HCLGa.fst
| fstrmsymbols $dir/disambig_tid.list | fstrmepslocal | fstminimizeencoded > $dir/HCLGa.fst
fstisstochastic $dir/HCLGa.fst || echo "HCLGa is not stochastic"

Просмотреть файл

@ -94,17 +94,18 @@ fi
fstisstochastic $clg || echo "warning: CLG not stochastic."
make-ilabel-transducer --write-disambig-syms=$dir/disambig_ilabels_remapped.list $dir/ilabels $tree $model $dir/ilabels.remapped > $dir/ilabel_map.fst
make-ilabel-transducer --write-disambig-syms=$dir/disambig_ilabels_remapped.list \
$dir/ilabels $tree $model $dir/ilabels.remapped > $dir/ilabel_map.fst
# Reduce size of CLG by remapping symbols...
fstcompose $dir/ilabel_map.fst $clg | fstdeterminizestar --use-log=true \
| fstminimizeencoded > $dir/CLG2.fst
make-h-transducer --disambig-syms-out=$dir/disambig_tstate.list \
make-h-transducer --disambig-syms-out=$dir/disambig_tid.list \
--transition-scale=$tscale $dir/ilabels.remapped $tree $model > $dir/Ha.fst
fsttablecompose $dir/Ha.fst $dir/CLG2.fst | fstdeterminizestar --use-log=true \
| fstrmsymbols $dir/disambig_tstate.list | fstrmepslocal | fstminimizeencoded > $dir/HCLGa.fst
| fstrmsymbols $dir/disambig_tid.list | fstrmepslocal | fstminimizeencoded > $dir/HCLGa.fst
fstisstochastic $dir/HCLGa.fst || echo "HCLGa is not stochastic"