Blame view

egs/wsj/s5/steps/lmrescore.sh 6.22 KB
8dcb6dfcb   Yannick Estève   first commit
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
  #!/bin/bash
  
  set -e -o pipefail
  
  # Begin configuration section.
  mode=4  # mode can be 1 through 5.  They should all give roughly similar results.
          # See the comments in the case statement for more details.
  cmd=run.pl
  skip_scoring=false
  self_loop_scale=0.1  # only matters for mode 4.
  acoustic_scale=0.1   # only matters for mode 5.
  # End configuration section.
  
  echo "$0 $@"  # Print the command line for logging
  
  . ./utils/parse_options.sh
  
  if [ $# != 5 ]; then
     echo "Do language model rescoring of lattices (remove old LM, add new LM)"
     echo "Usage: steps/lmrescore.sh [options] <old-lang-dir> <new-lang-dir> <data-dir> <input-decode-dir> <output-decode-dir>"
     echo "Ooptions:"
     echo " --cmd   <cmd-string>       # How to run commands (e.g. run.pl, queue.pl)"
     echo " --mode  (1|2|3|4|5)        # Mode of LM rescoring to use (default: 4)."
     echo "                            # These should give very similar results."
     echo " --self-loop-scale  <scale> # Self-loop-scale, only relevant in mode 4."
     echo "                            # Default: 0.1."
     echo " --acoustic-scale  <scale>  # Acoustic scale, only relevant in mode 5."
     echo "                            # Default: 0.1."
     exit 1;
  fi
  
  [ -f path.sh ] && . ./path.sh;
  
  oldlang=$1
  newlang=$2
  data=$3
  indir=$4
  outdir=$5
  
  oldlm=$oldlang/G.fst
  newlm=$newlang/G.fst
  ! cmp $oldlang/words.txt $newlang/words.txt && echo "Warning: vocabularies may be incompatible."
  [ ! -f $oldlm ] && echo Missing file $oldlm && exit 1;
  [ ! -f $newlm ] && echo Missing file $newlm && exit 1;
  ! ls $indir/lat.*.gz >/dev/null && echo "No lattices input directory $indir" && exit 1;
  
  if ! cmp -s $oldlang/words.txt $newlang/words.txt; then
    echo "$0: $oldlang/words.txt and $newlang/words.txt differ: make sure you know what you are doing.";
  fi
  
  oldlmcommand="fstproject --project_output=true $oldlm |"
  newlmcommand="fstproject --project_output=true $newlm |"
  
  mkdir -p $outdir/log
  
  phi=`grep -w '#0' $newlang/words.txt | awk '{print $2}'`
  
  if [ "$mode" == 4 ]; then
    # we have to prepare $outdir/Ldet.fst in this case: determinized
    # lexicon (determinized on phones), with disambig syms removed.
    # take L_disambig.fst; get rid of transition with "#0 #0" on it; determinize
    # with epsilon removal; remove disambiguation symbols.
    fstprint $newlang/L_disambig.fst | awk '{if($4 != '$phi'){print;}}' | fstcompile | \
      fstdeterminizestar | fstrmsymbols $newlang/phones/disambig.int >$outdir/Ldet.fst || exit 1;
  fi
  
  nj=`cat $indir/num_jobs` || exit 1;
  cp $indir/num_jobs $outdir
  
  
  #for lat in $indir/lat.*.gz; do
  #  number=`basename $lat | cut -d. -f2`;
  #  newlat=$outdir/`basename $lat`
  
  case "$mode" in
    1) # 1 is inexact, it's the original way of doing it.
      $cmd JOB=1:$nj $outdir/log/rescorelm.JOB.log \
        lattice-lmrescore --lm-scale=-1.0 "ark:gunzip -c $indir/lat.JOB.gz|" "$oldlmcommand" ark:-  \| \
        lattice-lmrescore --lm-scale=1.0 ark:- "$newlmcommand" "ark,t:|gzip -c>$outdir/lat.JOB.gz" \
        || exit 1;
      ;;
    2)  # 2 is equivalent to 1, but using more basic operations, combined.
      $cmd JOB=1:$nj $outdir/log/rescorelm.JOB.log \
        gunzip -c $indir/lat.JOB.gz \| \
        lattice-scale --acoustic-scale=-1 --lm-scale=-1 ark:- ark:- \| \
        lattice-compose ark:- "fstproject --project_output=true $oldlm |" ark:- \| \
        lattice-determinize ark:- ark:- \| \
        lattice-scale --acoustic-scale=-1 --lm-scale=-1 ark:- ark:- \| \
        lattice-compose ark:- "fstproject --project_output=true $newlm |" ark:- \| \
        lattice-determinize ark:- ark:- \| \
        gzip -c \>$outdir/lat.JOB.gz || exit 1;
      ;;
    3) # 3 is "exact" in that we remove the old LM scores accepting any path
       # through G.fst (which is what we want as that happened in lattice
       # generation), but we add the new one with "phi matcher", only taking
       # backoff arcs if an explicit arc did not exist.
      $cmd JOB=1:$nj $outdir/log/rescorelm.JOB.log \
        gunzip -c $indir/lat.JOB.gz \| \
        lattice-scale --acoustic-scale=-1 --lm-scale=-1 ark:- ark:- \| \
        lattice-compose ark:- "fstproject --project_output=true $oldlm |" ark:- \| \
        lattice-determinize ark:- ark:- \| \
        lattice-scale --acoustic-scale=-1 --lm-scale=-1 ark:- ark:- \| \
        lattice-compose --phi-label=$phi ark:- $newlm ark:- \| \
        lattice-determinize ark:- ark:- \| \
        gzip -c \>$outdir/lat.JOB.gz || exit 1;
      ;;
    4) # 4 is also exact (like 3), but instead of subtracting the old LM-scores,
       # it removes the old graph scores entirely and adds in the lexicon,
       # grammar and transition weights.
      mdl=`dirname $indir`/final.mdl
      [ ! -f $mdl ] && echo No such model $mdl && exit 1;
      [[ -f `dirname $indir`/frame_subsampling_factor && "$self_loop_scale" == 0.1 ]] && \
        echo "$0: WARNING: chain models need '--self-loop-scale 1.0'";
      $cmd JOB=1:$nj $outdir/log/rescorelm.JOB.log \
        gunzip -c $indir/lat.JOB.gz \| \
        lattice-scale --lm-scale=0.0 ark:- ark:- \| \
        lattice-to-phone-lattice $mdl ark:- ark:- \| \
        lattice-compose ark:- $outdir/Ldet.fst ark:- \| \
        lattice-determinize ark:- ark:- \| \
        lattice-compose --phi-label=$phi ark:- $newlm ark:- \| \
        lattice-add-trans-probs --transition-scale=1.0 --self-loop-scale=$self_loop_scale \
        $mdl ark:- ark:- \| \
        gzip -c \>$outdir/lat.JOB.gz  || exit 1;
      ;;
    5) # Mode 5 uses the binary lattice-lmrescore-pruned to do the LM rescoring
      # within a single program.  There are options for pruning, but these won't
      # normally need to be modified; the pruned aspect is more necessary for
      # RNNLM rescoring or when the lattices are extremely deep.
  
      [[ -f `dirname $indir`/frame_subsampling_factor && "$acoustic_scale" == 0.1 ]] && \
        echo "$0: WARNING: chain models need '--acoustic-scale 1.0'";
  
      $cmd JOB=1:$nj $outdir/log/rescorelm.JOB.log \
        lattice-lmrescore-pruned --acoustic-scale=$acoustic_scale "$oldlm" "$newlm" \
        "ark:gunzip -c $indir/lat.JOB.gz|" "ark:|gzip -c >$outdir/lat.JOB.gz" || exit 1;
      ;;
  esac
  
  rm $outdir/Ldet.fst 2>/dev/null || true
  
  if ! $skip_scoring ; then
    [ ! -x local/score.sh ] && \
      echo "Not scoring because local/score.sh does not exist or not executable." && exit 1;
    local/score.sh --cmd "$cmd" $data $newlang $outdir
  else
    echo "Not scoring because requested so..."
  fi
  
  exit 0;