def syllabifyTextgrids(tgPath, islePath):
isleDict = isletool.LexicalTool(islePath)
outputPath = join(tgPath, "syllabifiedTGs")
utils.makeDir(outputPath)
skipLabelList = ["<VOCNOISE>", "xx", "<SIL>", "{B_TRANS}", '{E_TRANS}']
for fn in utils.findFiles(tgPath, filterExt=".TextGrid"):
if os.path.exists(join(outputPath, fn)):
continue
tg = tgio.openTextgrid(join(tgPath, fn))
syllableTG = praattools.syllabifyTextgrid(isleDict,
tg,
"words",
"phones",
skipLabelList=skipLabelList)
outputTG = tgio.Textgrid()
outputTG.addTier(tg.tierDict["words"])
outputTG.addTier(tg.tierDict["phones"])
# outputTG.addTier(syllableTG.tierDict["syllable"])
outputTG.addTier(syllableTG.tierDict["tonic"])
outputTG.save(join(outputPath, fn))
def manualPhoneCount(tgInfoPath, isleFN, outputPath, skipList=None):
if skipList is None:
skipList = []
utils.makeDir(outputPath)
isleDict = isletool.LexicalTool(isleFN)
existFNList = utils.findFiles(outputPath, filterPaths=".txt")
for fn in utils.findFiles(tgInfoPath, filterExt=".txt",
skipIfNameInList=existFNList):
if os.path.exists(join(outputPath, fn)):
continue
print(fn)
dataList = utils.openCSV(tgInfoPath, fn)
dataList = [row[2] for row in dataList] # start, stop, tmpLabel
outputList = []
for tmpLabel in dataList:
if tmpLabel not in skipList:
syllableCount, phoneCount = isletool.getNumPhones(isleDict,
tmpLabel,
maxFlag=True)
else:
syllableCount, phoneCount = 0, 0
outputList.append("%d,%d" % (syllableCount, phoneCount))
outputTxt = "\n".join(outputList)
with open(join(outputPath, fn), "w") as fd:
fd.write(outputTxt)
def createSyllabifiedTextgrid(names_audio, text_grid):
isleDict = isletool.LexicalTool()
tg = tgio.openTextgrid(text_grid)
syllableTG = praattools.syllabifyTextgrid(isleDict,
tg,
"words",
"phones",
skipLabelList=[
"",
])
tg.addTier(syllableTG.tierDict["syllable"])
tg.addTier(syllableTG.tierDict["tonicSyllable"])
tg.addTier(syllableTG.tierDict["tonicVowel"])
tg.save('../mavid-scripts/files/wav_recordedNames_syllables_test.TextGrid')
return
# -*- coding: utf-8 -*-
"""
Created on Mon Aug 26 11:20:12 2019
@author: john.cheng
"""
import random
from os.path import join
from pysle import isletool
from pysle import pronunciationtools
root = join(".", "files")
isleDict = isletool.LexicalTool(join(root, 'D:\Aeo_test\speech\ISLEdict.txt'))
def printOutMatches(matchStr,
numSyllables=None,
wordInitial='ok',
wordFinal='ok',
spanSyllable='ok',
stressedSyllable='ok',
multiword='ok',
numMatches=None,
matchList=None,
pos=None):
if matchList is None:
matchList = isleDict.search(matchStr, numSyllables, wordInitial,
wordFinal, spanSyllable, stressedSyllable,
multiword, pos)
#encoding: utf-8
'''
Examples of how to use pysle's pronunciationtools code
'''
from os.path import join
from pysle import isletool
from pysle import pronunciationtools
root = join(".", "files")
isleDict = isletool.LexicalTool(join(root, 'ISLEdict_sample.txt'))
# In the first example we determine the syllabification of a word,
# as it was said. (Of course, this is just an estimate)
print('-' * 50)
searchWord = 'another'
anotherPhoneList = ['n', '@', 'th', 'r']
isleWordList = isleDict.lookup(searchWord)
returnList = pronunciationtools.findBestSyllabification(
isleDict, searchWord, anotherPhoneList)
(stressedSyllable, stressedPhone, syllableList, syllabification,
stressedSyllableIndexList, stressedPhoneIndexList,
flattenedStressIndexList) = returnList
print(searchWord)
print(anotherPhoneList)
print(stressedSyllableIndexList) # We can see the first syllable was elided
print(stressedPhoneIndexList)
print(flattenedStressIndexList)
print(syllableList)
#encoding: utf-8
'''
Created on July 08, 2016
@author: tmahrt
Basic examples of common usage.
'''
import random
from pysle import isletool
tmpPath = r"C:\Users\Tim\Dropbox\workspace\pysle\test\ISLEdict.txt"
isleDict = isletool.LexicalTool(tmpPath)
def printOutMatches(matchStr, numSyllables=None, wordInitial='ok',
wordFinal='ok', spanSyllable='ok', stressedSyllable='ok',
multiword='ok', numMatches=None, matchList=None):
if matchList is None:
matchList = isleDict.search(matchStr, numSyllables, wordInitial,
wordFinal, spanSyllable, stressedSyllable,
multiword)
else:
matchList = isletool.search(matchList, matchStr, numSyllables, wordInitial,
wordFinal, spanSyllable, stressedSyllable,
multiword)
if numMatches is not None and len(matchList) > numMatches:
random.shuffle(matchList)
'''
from os.path import join
from praatio import tgio
from pysle import isletool
from pysle import praattools
path = join('.', 'files')
path = "/Users/tmahrt/Dropbox/workspace/pysle/test/files"
tg = tgio.openTextGrid(join(path, "pumpkins.TextGrid"))
# Needs the full path to the file
islevPath = '/Users/tmahrt/Dropbox/workspace/pysle/test/islev2.txt'
isleDict = isletool.LexicalTool(islevPath)
# Get the syllabification tiers and add it to the textgrid
syllableTG = praattools.syllabifyTextgrid(isleDict, tg, "word", "phone",
skipLabelList=["",])
tg.addTier(syllableTG.tierDict["syllable"])
tg.addTier(syllableTG.tierDict["tonicSyllable"])
tg.addTier(syllableTG.tierDict["tonicVowel"])
tg.save(join(path, "pumpkins_with_syllables.TextGrid"))
- 10 syllables?
- Alternating stressed/unstressed?
3. Print line
'''
import random
import time
import string
import markovify
from os.path import join
from pysle import isletool
from pysle import pronunciationtools
root = join(".", "files")
isleDict = isletool.LexicalTool(
join(root, '/home/jay/Dropbox/19-20/PoetryGen/ISLEdict.txt'))
# Set your text corpus here
with open("/home/jay/Dropbox/19-20/PoetryGen/shelley.txt") as f:
shelley = f.read()
with open("/home/jay/Dropbox/19-20/PoetryGen/mobydick.txt") as f:
mobydick = f.read()
with open("/home/jay/Dropbox/19-20/PoetryGen/witchcraft.txt") as f:
witch = f.read()
textModelShelley = markovify.Text(shelley, state_size=2)
textModelDick = markovify.Text(mobydick, state_size=2)
textModelWitch = markovify.Text(witch, state_size=2)
backoff = Backoff(['fas-Arab', 'rus-Cyrl'])
backoff.transliterate('Привет дорогой друг пидор')
backoff.transliterate('queen')
backoff.transliterate('中文')
backoff.transliterate('سلام شادی من')
backoff.transliterate('ملکه')
from pysle import isletool
a = isletool.LexicalTool('ISLEdict.txt')
sentence = "do you want another pumpkinseed"
phoneList = isletool.transcribe(a, sentence, 'longest')
print(phoneList)
from phonemizer.phonemize import phonemize
phonemize("hello", language='en-us', backend='espeak')
phonemize("hello my queen", language='en-us', backend='espeak')
phonemize("ich will", language='de', backend='espeak')
phonemize("bonjour le monde", language='fr-fr', backend='espeak')
phonemize("konnichiwa", language='japanese', backend='espeak')
def get_data(seed=42, test_size=0.20, verbose=0, maxlen_x=None, maxlen_y=None, blacklist='()0123456789%.?"-_', max_phonemes=np.inf, max_chars=np.inf, phon_sep='', unique_graphemes=False, unique_phonemes=True):
"""Process ISLEDICT pronounciation dictionary to return unique phonemes two graphemes"""
path = download_data_maybe()
# load data
isleDict = isletool.LexicalTool(path)
X = []
y = []
for phrase in isleDict.data.keys():
for pronounciation in zip(*isleDict.lookup(phrase)):
xx = []
for syllableList, stressedSyllableList, stressedPhoneList in pronounciation:
xx += list(itertools.chain(*syllableList))
y.append(phon_sep.join(xx))
X.append(phrase)
if verbose: print('loaded entries {}'.format(len(X)))
# filter out duplicate X's
if unique_phonemes:
y, X = zip(*dict(zip(y, X)).items())
if verbose: print('removed duplicate phonemes leaving {}'.format(len(X)))
# filter out duplicates Y's
if unique_graphemes:
X, y = zip(*dict(zip(X, y)).items())
if verbose: print('removed duplicate graphemes leaving {}'.format(len(X)))
# split data (we must set asside test data before cleanign so it's always the same)
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=test_size, random_state=seed)
# filter out duplicate entries like 'HOUSE(2) or multi words CAT-DOG and CAT_DOG'
p = re.compile('[%s]' % (re.escape(blacklist)))
X_train, y_train = zip(*[(x, y) for x, y in zip(X_train, y_train) if not bool(p.findall(x))])
X_test, y_test = zip(*[(x, y) for x, y in zip(X_test, y_test) if not bool(p.findall(x))])
if verbose:
print('removed blacklisted entries leaving {}'.format(len(X_train) + len(X_test)))
# filter out complex entries if needed
before_x = len(y_train)
X_train, y_train = zip(*[(x, y) for x, y in zip(X_train, y_train) if len(y) <= max_phonemes and len(x) <= max_chars])
X_test, y_test = zip(*[(x, y) for x, y in zip(X_test, y_test) if len(y) <= max_phonemes and len(x) <= max_chars])
if verbose:
print('restricted to less than {} phonemes leaving {} entries or {:2.2f}%'.format(max_phonemes, len(X_train) + len(X_test), len(X_train)/before_x*100))
# FIXME it's slow in the next few lines
# encode x and y and pad them
xtable = CharacterTable()
xtable.fit(X_test + X_train)
if maxlen_x:
xtable.maxlen = maxlen_x
X_train = xtable.encode(X_train)
X_test = xtable.encode(X_test)
ytable = CharacterTable()
ytable.fit(y_test + y_train)
if maxlen_y:
ytable.maxlen = maxlen_y
y_train = ytable.encode(y_train)
y_test = ytable.encode(y_test)
if verbose:
print('X_train shape:', X_train.shape)
print('X_test shape:', X_test.shape)
print('y_train shape:', y_train.shape)
print('y_test shape:', y_test.shape)
return (X_train, y_train), (X_test, y_test), (xtable, ytable)
