import logging
import math
import os
import shutil
import subprocess
import sys
import tempfile
from paths import get_binary
from metasentence import MetaSentence
MKGRAPH_PATH = get_binary("ext/mkgraph")
def make_bigram_lm_fst(word_sequences, **kwargs):
'''
Use the given token sequence to make a bigram language model
in OpenFST plain text format.
When the "conservative" flag is set, an [oov] is interleaved
between successive words.
When the "disfluency" flag is set, a small set of disfluencies is
interleaved between successive words
`Word sequence` is a list of lists, each valid as a start
'''
if len(word_sequences) == 0 or type(word_sequences[0]) != list:
word_sequences = [word_sequences]
conservative = kwargs['conservative'] if 'conservative' in kwargs else False
import subprocess
from paths import get_binary
FFMPEG = get_binary("ffmpeg")
def to_wav(infile, outfile):
'''
Use FFMPEG to convert a media file to a wav file
'''
return subprocess.call([
FFMPEG, '-loglevel', 'panic', '-y', '-i', infile, '-ac', '1', '-ar',
'8000', '-acodec', 'pcm_s16le', outfile
])
import subprocess
from paths import get_binary
FFMPEG = get_binary("ffmpeg")
def to_wav(path, R=8000, depth=16, nchannels=1, start=0):
'''
Use FFMPEG to convert a media file to a wav file with the given
sample format.
Returns an IO object so the results can be streamed.
'''
cmd = [FFMPEG,
'-ss', "%f" % (start),
'-i', path,
'-loglevel', 'panic',
'-ss', "%f" % (start),
'-vn',
'-ar', str(R),
'-ac', str(nchannels),
'-f', 'wav',
'-acodec', 'pcm_s16le',
'-']
p = subprocess.Popen(cmd, stdout=subprocess.PIPE, stderr=open('/dev/null', 'w'))
return p.stdout
import logging
import math
import os
import shutil
import subprocess
import sys
import tempfile
from paths import get_binary
from metasentence import MetaSentence
MKGRAPH_PATH = get_binary("ext/mkgraph")
def make_bigram_lm_fst(word_sequence):
'''
Use the given token sequence to make a bigram language model
in OpenFST plain text format.
'''
word_sequence = ['[oov]', '[oov]'] + word_sequence + ['[oov]']
bigrams = {}
prev_word = word_sequence[0]
for word in word_sequence[1:]:
bigrams.setdefault(prev_word, set()).add(word)
prev_word = word
node_ids = {}
def get_node_id(word):
node_id = node_ids.get(word, len(node_ids) + 1)
node_ids[word] = node_id
return node_id
