def _lexicon(self):
corpus_file = "{}/data/local/corpus.txt".format(self.tmp_folder)
texts = open(corpus_file).read()
words = sorted(set(texts.split()))
phones = [list(PhoneConverter.word2phone(item)) for item in words]
nonsilence_phones = sorted(
set([item for sublist in phones for item in sublist]))
outfile = "{}/data/local/dict/nonsilence_phones.txt".format(
self.tmp_folder)
content = "\n".join(nonsilence_phones) + "\n"
open(outfile, "w").write(content)
silence_phones = ["sil", "spn"]
outfile = "{}/data/local/dict/silence_phones.txt".format(
self.tmp_folder)
content = "\n".join(silence_phones) + "\n"
open(outfile, "w").write(content)
optional_silence = ["sil"]
outfile = "{}/data/local/dict/optional_silence.txt".format(
self.tmp_folder)
content = "\n".join(optional_silence) + "\n"
open(outfile, "w").write(content)
lexicon = ([
"{} {}".format(word, " ".join(phone))
for word, phone in zip(words, phones)
])
lexicon = ["!SIL sil", "<UNK> spn"] + lexicon
content = "\n".join(lexicon) + "\n"
outfile = "{}/data/local/dict/lexicon.txt".format(self.tmp_folder)
open(outfile, "w").write(content)
def _make_dictionary(self):
lines = open(
"{}/train/text".format(self.corpus_folder)).read().splitlines()[
:self.N_TRAIN]
phones = []
for line in lines:
fileid, word = line.split("|")
p = PhoneConverter.word2phone(word).split()
phones += p
phones = sorted(set(phones))
# create .dic files
lines = []
phone_units = []
for p in phones:
units = list(p)
phone_units += units
units = " ".join(units)
line = "{:20s}{}".format(p, units)
lines.append(line)
open("{}/etc/tmp.dic".format(self.tmp_folder), "w").write(
"\n".join(lines))
phone_units = sorted(set(phone_units))
phone_units.append("SIL")
open("{}/etc/tmp.phone".format(self.tmp_folder), "w").write(
"\n".join(phone_units))
def _make_cleaned_text(self):
in_file = "{}/train/text".format(self.corpus_folder)
out_file = "{}/etc/text".format(self.tmp_folder)
lines = open(in_file).read().splitlines()[:self.N_TRAIN]
output = []
for line in lines:
fileid, word = line.split("|")
phone = PhoneConverter.word2phone(word)
content = "<s> {} </s>".format(phone, fileid)
output.append(content)
content = "\n".join(output)
open(out_file, "w").write(content)