def divideIntoSentencesFromAnno(annotationURI):
'''
infer section/line timestamps from annotation-textgrid,
parse divison into sentences from Tier 'lines' and load its syllables corresponding by timestamps
'''
whichLevel = 5 # read lines (sentences) tier
annotationTokenList, annotationLinesListNoPauses = readNonEmptyTokensTextGrid(annotationURI, whichLevel, 0, -1)
whichLevel = 3 # read syllables as pinyin
syllablesList = TextGrid2WordList(annotationURI, whichLevel)
annotationTokenList, syllablesList = readNonEmptyTokensTextGrid(annotationURI, whichLevel, 0, -1)
syllablePointer = 0
listSentences = []
for currSentence in annotationLinesListNoPauses:
currSectionSyllables = []
currSentenceBegin = currSentence[0]
currSentenceEnd = currSentence[1]
while syllablesList[syllablePointer][0] < currSentenceBegin: # search for beginning
syllablePointer += 1
if not syllablesList[syllablePointer][0] == currSentenceBegin: # start has to be aligned
sys.exit("no syllable starting at sentence start at {} ".format(currSentenceBegin) )
fromSyllableIdx = syllablesList[syllablePointer][3]
while syllablePointer < len(syllablesList) and float(syllablesList[syllablePointer][1]) <= currSentenceEnd: # syllables in currSentence
isEndOfSentence, syllableTxt = stripPunctuationSings(syllablesList[syllablePointer][2])
currSyllable = SyllableJingju(syllableTxt, -1)
currSyllable.setDurationInMinUnit(1)
currSectionSyllables.append(currSyllable)
syllablePointer += 1
if not syllablesList[syllablePointer-1][1] == currSentenceEnd: # end has to be aligned
sys.exit("no syllable ending at sentence end at {} ".format(currSentenceEnd) )
toSyllableIdx = syllablesList[syllablePointer-1][3]
listSentences.append(( currSentenceBegin, currSentenceEnd, fromSyllableIdx, toSyllableIdx, currSectionSyllables))
return listSentences
def divideIntoSentencesFromAnnoOld(annotationURI):
'''
infer section/line timestamps from annotation-textgrid,
use punctuation as marker for sentence ends
@deprecated
'''
# whichLevel = 5 # line
whichLevel = 3 # pinyin
annotationTokenList, annotationTokenListNoPauses = readNonEmptyTokensTextGrid(annotationURI, whichLevel, 0, -1)
currSectionSyllables = []
listSentences = []
i = 0
currSectionStartTime = annotationTokenListNoPauses[i][0]
fromSyllable = annotationTokenListNoPauses[i][3]
for i in range(len(annotationTokenListNoPauses)):
token = annotationTokenListNoPauses[i]
isEndOfSentence, token[2] = stripPunctuationSings(token[2])
if isEndOfSentence:
currSyllable = SyllableJingju(token[2], -1)
currSyllable.setDurationInMinUnit(1)
currSectionSyllables.append(currSyllable)
currSectionEndTime = token[1]
toSyllable = token[3]
listSentences.append(( currSectionStartTime, currSectionEndTime, fromSyllable, toSyllable, currSectionSyllables))
# start next section
currSectionSyllables = []
if i != len(annotationTokenListNoPauses)-1:
currSectionStartTime = annotationTokenListNoPauses[i+1][0]
fromSyllable = annotationTokenListNoPauses[i+1][3]
else: # syllable not at end of sentence
currSyllable = SyllableJingju(token[2], -1)
currSyllable.setDurationInMinUnit(1)
currSectionSyllables.append(currSyllable)
return listSentences
def createSyllables(annotationURI, fromSyllable, toSyllable):
'''
@param refSyllableDuration: its value does not matter. important is that all syllables are assigned same relative duration.
create Syllables, assign their durations in refSyllableDuration
@return: lyrics - created lyrics oboject
'''
listSyllables = []
annotationTokenList, annotationTokenListNoPauses = readNonEmptyTokensTextGrid(annotationURI, 3, fromSyllable, toSyllable)
for tsAndSyll in annotationTokenListNoPauses:
currSyllable = SyllableJingju(tsAndSyll[2], -1)
currSyllable.setDurationInMinUnit(1)
listSyllables.append(currSyllable)
return listSyllables
