Blame view
egs/wsj/s5/utils/make_lexicon_fst.pl
5.79 KB
8dcb6dfcb 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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 |
#!/usr/bin/env perl # THIS SCRIPT IS DEPRECATED AND WILL BE REMOVED. See # utils/lang/make_lexicon_fst.py which is the python-based replacement. use warnings; #sed replacement for -w perl parameter # Copyright 2010-2011 Microsoft Corporation # 2013 Johns Hopkins University (author: Daniel Povey) # Licensed under the Apache License, Version 2.0 (the "License"); # you may not use this file except in compliance with the License. # You may obtain a copy of the License at # # http://www.apache.org/licenses/LICENSE-2.0 # # THIS CODE IS PROVIDED *AS IS* BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY # KIND, EITHER EXPRESS OR IMPLIED, INCLUDING WITHOUT LIMITATION ANY IMPLIED # WARRANTIES OR CONDITIONS OF TITLE, FITNESS FOR A PARTICULAR PURPOSE, # MERCHANTABLITY OR NON-INFRINGEMENT. # See the Apache 2 License for the specific language governing permissions and # limitations under the License. # makes lexicon FST, in text form, from lexicon (pronunciation probabilities optional). $pron_probs = 0; if ((@ARGV > 0) && ($ARGV[0] eq "--pron-probs")) { $pron_probs = 1; shift @ARGV; } if (@ARGV != 1 && @ARGV != 3 && @ARGV != 4) { print STDERR "Usage: make_lexicon_fst.pl [--pron-probs] lexicon.txt [silprob silphone [sil_disambig_sym]] >lexiconfst.txt "; print STDERR "Creates a lexicon FST that transduces phones to words, and may allow optional silence. "; print STDERR "Note: ordinarily, each line of lexicon.txt is: "; print STDERR " word phone1 phone2 ... phoneN; "; print STDERR "if the --pron-probs option is used, each line is: "; print STDERR " word pronunciation-probability phone1 phone2 ... phoneN. "; print STDERR "The probability 'prob' will typically be between zero and one, and note that "; print STDERR "it's generally helpful to normalize so the largest one for each word is 1.0, but "; print STDERR "this is your responsibility. "; print STDERR "The silence disambiguation symbol, e.g. something like #5, is used only "; print STDERR "when creating a lexicon with disambiguation symbols, e.g. L_disambig.fst, "; print STDERR "and was introduced to fix a particular case of non-determinism of decoding graphs. "; exit(1); } $lexfn = shift @ARGV; if (@ARGV == 0) { $silprob = 0.0; } elsif (@ARGV == 2) { ($silprob,$silphone) = @ARGV; } else { ($silprob,$silphone,$sildisambig) = @ARGV; } if ($silprob != 0.0) { $silprob < 1.0 || die "Sil prob cannot be >= 1.0"; $silcost = -log($silprob); $nosilcost = -log(1.0 - $silprob); } open(L, "<$lexfn") || die "Error opening lexicon $lexfn"; if ( $silprob == 0.0 ) { # No optional silences: just have one (loop+final) state which is numbered zero. $loopstate = 0; $nextstate = 1; # next unallocated state. while (<L>) { @A = split(" ", $_); @A == 0 && die "Empty lexicon line."; foreach $a (@A) { if ($a eq "<eps>") { die "Bad lexicon line $_ (<eps> is forbidden)"; } } $w = shift @A; if (! $pron_probs) { $pron_cost = 0.0; } else { $pron_prob = shift @A; if (! defined $pron_prob || !($pron_prob > 0.0 && $pron_prob <= 1.0)) { die "Bad pronunciation probability in line $_"; } $pron_cost = -log($pron_prob); } if ($pron_cost != 0.0) { $pron_cost_string = "\t$pron_cost"; } else { $pron_cost_string = ""; } $s = $loopstate; $word_or_eps = $w; while (@A > 0) { $p = shift @A; if (@A > 0) { $ns = $nextstate++; } else { $ns = $loopstate; } print "$s\t$ns\t$p\t$word_or_eps$pron_cost_string "; $word_or_eps = "<eps>"; $pron_cost_string = ""; # so we only print it on the first arc of the word. $s = $ns; } } print "$loopstate\t0 "; # final-cost. } else { # have silence probs. $startstate = 0; $loopstate = 1; $silstate = 2; # state from where we go to loopstate after emitting silence. print "$startstate\t$loopstate\t<eps>\t<eps>\t$nosilcost "; # no silence. if (!defined $sildisambig) { print "$startstate\t$loopstate\t$silphone\t<eps>\t$silcost "; # silence. print "$silstate\t$loopstate\t$silphone\t<eps> "; # no cost. $nextstate = 3; } else { $disambigstate = 3; $nextstate = 4; print "$startstate\t$disambigstate\t$silphone\t<eps>\t$silcost "; # silence. print "$silstate\t$disambigstate\t$silphone\t<eps> "; # no cost. print "$disambigstate\t$loopstate\t$sildisambig\t<eps> "; # silence disambiguation symbol. } while (<L>) { @A = split(" ", $_); $w = shift @A; if (! $pron_probs) { $pron_cost = 0.0; } else { $pron_prob = shift @A; if (! defined $pron_prob || !($pron_prob > 0.0 && $pron_prob <= 1.0)) { die "Bad pronunciation probability in line $_"; } $pron_cost = -log($pron_prob); } if ($pron_cost != 0.0) { $pron_cost_string = "\t$pron_cost"; } else { $pron_cost_string = ""; } $s = $loopstate; $word_or_eps = $w; while (@A > 0) { $p = shift @A; if (@A > 0) { $ns = $nextstate++; print "$s\t$ns\t$p\t$word_or_eps$pron_cost_string "; $word_or_eps = "<eps>"; $pron_cost_string = ""; $pron_cost = 0.0; # so we only print it the 1st time. $s = $ns; } elsif (!defined($silphone) || $p ne $silphone) { # This is non-deterministic but relatively compact, # and avoids epsilons. $local_nosilcost = $nosilcost + $pron_cost; $local_silcost = $silcost + $pron_cost; print "$s\t$loopstate\t$p\t$word_or_eps\t$local_nosilcost "; print "$s\t$silstate\t$p\t$word_or_eps\t$local_silcost "; } else { # no point putting opt-sil after silence word. print "$s\t$loopstate\t$p\t$word_or_eps$pron_cost_string "; } } } print "$loopstate\t0 "; # final-cost. } |