farsdat_prepare_dict.sh
3.04 KB
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
#!/bin/bash
# Copyright 2013 (Authors: Daniel Povey, Bagher BabaAli)
# 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.
# Call this script from one level above, e.g. from the s3/ directory. It puts
# its output in data/local/.
# The parts of the output of this that will be needed are
# [in data/local/dict/ ]
# lexicon.txt
# extra_questions.txt
# nonsilence_phones.txt
# optional_silence.txt
# silence_phones.txt
# run this from ../
srcdir=data/train
dir=data/local/dict
lmdir=data/local/nist_lm
tmpdir=data/local/lm_tmp
mkdir -p $dir $lmdir $tmpdir
[ -f path.sh ] && . ./path.sh
#(1) Dictionary preparation:
# Make phones symbol-table (adding in silence and verbal and non-verbal noises at this point).
# silence phones, one per line.
echo sil > $dir/silence_phones.txt
echo sil > $dir/optional_silence.txt
# nonsilence phones; on each line is a list of phones that correspond
# really to the same base phone.
# Create the lexicon, which is just an identity mapping
cut -d' ' -f2- $srcdir/text | tr ' ' '\n' | sort -u | sed '/^$/d' > $dir/phones.txt
paste $dir/phones.txt $dir/phones.txt > $dir/lexicon.txt || exit 1;
grep -v -F -f $dir/silence_phones.txt $dir/phones.txt > $dir/nonsilence_phones.txt
# A few extra questions that will be added to those obtained by automatically clustering
# the "real" phones. These ask about stress; there's also one for silence.
cat $dir/silence_phones.txt| awk '{printf("%s ", $1);} END{printf "\n";}' > $dir/extra_questions.txt || exit 1;
cat $dir/nonsilence_phones.txt | perl -e 'while(<>){ foreach $p (split(" ", $_)) {
$p =~ m:^([^\d]+)(\d*)$: || die "Bad phone $_"; $q{$2} .= "$p "; } } foreach $l (values %q) {print "$l\n";}' \
>> $dir/extra_questions.txt || exit 1;
# (2) Create the phone bigram LM
if [ -z $IRSTLM ] ; then
export IRSTLM=$KALDI_ROOT/tools/irstlm/
fi
export PATH=${PATH}:$IRSTLM/bin
if ! command -v prune-lm >/dev/null 2>&1 ; then
echo "$0: Error: the IRSTLM is not available or compiled" >&2
echo "$0: Error: We used to install it by default, but." >&2
echo "$0: Error: this is no longer the case." >&2
echo "$0: Error: To install it, go to $KALDI_ROOT/tools" >&2
echo "$0: Error: and run extras/install_irstlm.sh" >&2
exit 1
fi
cut -d' ' -f2- $srcdir/text | sed -e 's:^:<s> :' -e 's:$: </s>:' \
> $srcdir/lm_train
build-lm.sh -i $srcdir/lm_train -n 2 \
-o $tmpdir/lm_phone_bg.ilm.gz
compile-lm $tmpdir/lm_phone_bg.ilm.gz -t=yes /dev/stdout | \
grep -v unk | gzip -c > $lmdir/lm_phone_bg.arpa.gz
echo "Dictionary & language model preparation succeeded"