def save_intervalframe_to_textgrid(framedict, filepath, encoding='utf-8'):
"""Write a dict of IntervalFrames in a textgrid-File.
Arguments:
framedict -- Dictionary of dataframes. The keys become tier
names in the textgrid file
filepath -- Path + filename of the file to be written.
Keyword arguments:
encoding: character encoding to save textgrid file
"""
if len(framedict) < 1:
print "invalid data!"
return
mytextgrid = tgt.TextGrid()
for tier_name in framedict.keys():
newtier = framedict[tier_name]
if len(newtier.columns) == 3:
mytier = tgt.IntervalTier(name=tier_name)
for row in newtier.index:
myinterval = tgt.Interval(newtier[newtier.columns[0]][row],
newtier[newtier.columns[1]][row],
newtier[newtier.columns[2]][row])
mytier.add_interval(myinterval)
elif len(newtier.columns) == 2:
mytier = tgt.PointTier(name=tier_name)
for row in newtier.index:
mypoint = tgt.Point(newtier[newtier.columns[0]][row],
newtier[newtier.columns[1]][row])
mytier.add_point(mypoint)
mytextgrid.add_tier(mytier)
tgt.write_to_file(mytextgrid, filepath, encoding=encoding, format="long")
def main():
# Parse the command-line arguments.
args = parse_arguments()
tg_path = args['tg_path']
offset_start = args['offset_start']
offset_end = args['offset_end']
outpath = args['outpath']
# Read the TextGrid
tg = tgt.read_textgrid(tg_path)
tg_part = tgt.TextGrid()
if offset_start is None and offset_end is None:
raise Exception(
'At least one of offset_start and offset_end must be specified.')
elif offset_start is None:
offset_start = tg.start_time
elif offset_end is None:
offset_end = tg.end_time
for tr in tg:
intr_part = tr.get_annotations_between_timepoints(
offset_start, offset_end)
tier_part = tgt.IntervalTier(name=tr.name,
start_time=tr.start_time,
end_time=tr.end_time,
objects=intr_part)
tg_part.add_tier(tier_part)
if outpath is None:
tg_dirname, tg_filename = os.path.split(tg_path)
outpath = os.path.splitext(tg_filename)[0] + '_part.TextGrid'
tgt.write_to_file(tg_part, outpath)
def save_annotations(self, filename, tiers=['cycles', 'holds'],
filetype='textgrid', merge_holds=False):
"""Save annotations to file."""
if filetype not in ['textgrid', 'eaf', 'table']:
raise ValueError('Unsupported file type: {}'.format(filetype))
tg = tgt.TextGrid()
if 'holds' in tiers or merge_holds:
# holds = tgt.IntervalTier(name='holds')
# for start, end in self.holds:
# holds.add_interval(tgt.Interval(start, end, 'hold'))
if not merge_holds:
tg.add_tier(self.holds)
if 'cycles' in tiers:
if merge_holds:
segments_merged= tgt.IntervalTier(name='cycles')
tg.add_tier(self.merge_holds(self.segments, self.holds))
else:
tg.add_tier(self.segments)
if len(tg.tiers):
filetype = 'short' if filetype == 'textgrid' else filetype
tgt.write_to_file(tg, filename, format=filetype)
def save_intervalframe_to_textgrid(framedict, filepath, encoding='utf-8'):
"""Write a dict of IntervalFrames in a textgrid-File.
Arguments:
framedict -- Dictionary of dataframes. The keys become tier
names in the textgrid file
filepath -- Path + filename of the file to be written.
Keyword arguments:
encoding: character encoding to save textgrid file
"""
if len(framedict) < 1:
print "invalid data!"
return
mytextgrid = tgt.TextGrid()
for tier_name in framedict.keys():
newtier = framedict[tier_name]
if len(newtier.columns) == 3:
mytier = tgt.IntervalTier(name=tier_name)
for row in newtier.index:
myinterval = tgt.Interval(newtier[newtier.columns[0]][row],
newtier[newtier.columns[1]][row],
newtier[newtier.columns[2]][row])
mytier.add_interval(myinterval)
elif len(newtier.columns) == 2:
mytier = tgt.PointTier(name=tier_name)
for row in newtier.index:
mypoint = tgt.Point(newtier[newtier.columns[0]][row],
newtier[newtier.columns[1]][row])
mytier.add_point(mypoint)
mytextgrid.add_tier(mytier)
tgt.write_to_file(mytextgrid, filepath, encoding=encoding, format="long")
def add_lemmas(title, input1_path, output_path):
# Load textgrid
tg = tgt.read_textgrid(os.path.join(input1_path, title + '.TextGrid'))
tier_names = tg.get_tier_names()
# Load pos tier
pos_tier_name = [name for name in tier_names if 'pos' in name][0]
pos_tier = tg.get_tier_by_name(pos_tier_name)
# Load words tier
words_tier_name = [name for name in tier_names if 'words' in name][0]
words_tier = tg.get_tier_by_name(words_tier_name)
# Start empty lemmas tier
lemmas_tier = tgt.IntervalTier()
lemmas_tier_name = [name for name in tier_names
if 'words' in name][0].replace('words', 'lemmas')
lemmas_tier.name = lemmas_tier_name
# Generate lemma intervals
lemmas_intervals = [
tgt.Interval(w_interval.start_time, w_interval.end_time,
lemmatize_word(w_interval.text, pos_tier[i].text))
for i, w_interval in enumerate(words_tier)
]
# Add lemmas to tier
lemmas_tier.add_annotations(lemmas_intervals)
tg.add_tier(lemmas_tier)
tgt.write_to_file(tg,
os.path.join(output_path, title + '.TextGrid'),
format='short')
def main():
args = parse_arguments()
for input_path in args['input_files']:
#Parse the input file
annotation_root = etree.parse(input_path).getroot()
# Only do the conversion if neither the --list-features nor the --list-tracks
# command-line arguments have been passed.
if not (args['list_features'] or args['list_tracks']):
# Unless an output_dir is specified, save output files to the directory of the input file
input_dir, input_filename = os.path.split(input_path)
if input_filename[-4:] == '.xml':
output_filename = input_filename[:-4] + '.TextGrid'
else:
output_filename = input_filename + '.TextGrid'
if args['output_dir'] is not None:
output_path = os.path.join(args['output_dir'][0], output_filename)
else:
output_path = os.path.join(input_dir, output_filename)
textgrid = higgins2tg(annotation_root, args['features_included'], args['features_excluded'],
args['tracks_included'], args['tracks_excluded'])
tgt.write_to_file(textgrid, output_path, encoding=args['output_encoding'], format=args['textgrid_type'])
elif args['list_features']:
features_all = get_all_features(annotation_root)
sys.stdout.write('%s -- Features: ' % input_filename+ ', '.join(features_all) + '\n')
else:
tracks = get_tracks(annotation_root)
sys.stdout.write('%s -- Tracks: ' % input_filename + ', '.join(tracks) + '\n')
def add_onsets_rhymes(title, input_path, output_path):
# Load the textgrid
tg = tgt.read_textgrid(os.path.join(input_path, title + '.TextGrid'))
# Load name of all tiers
tier_names = tg.get_tier_names()
# Select a tier whose name contains 'syllables'
sylls_tier_name = [name for name in tier_names if 'sylls' in name][0]
sylls_tier = tg.get_tier_by_name(sylls_tier_name)
# Select a tier whose name contains 'phones'
phones_tier_name = [name for name in tier_names if 'phones' in name][0]
phones_tier = tg.get_tier_by_name(phones_tier_name)
# Start an empty tier for onset-rhymes
onset_rhyme_tier = tgt.IntervalTier()
onset_rhyme_tier_name = [name for name in tier_names
if 'words' in name][0].replace('words', 'OR')
onset_rhyme_tier.name = onset_rhyme_tier_name
onset_rhyme_intervals = []
for syll in sylls_tier._get_annotations():
#print(syll)
phs = phones_tier.get_annotations_between_timepoints(
syll.start_time, syll.end_time)
nucleus_index = calculate_nucleus_index(phs)
# If the first phone contains a number then it means the whole syll has no onset, so we only add a rhyme
if nucleus_index == 0:
onset_rhyme_intervals.append(
tgt.Interval(syll.start_time, syll.end_time, 'R'))
# If the onset is present add onset and rhyme intervals
else:
onset_rhyme_intervals.append(
tgt.Interval(syll.start_time, phs[nucleus_index - 1].end_time,
'O'))
onset_rhyme_intervals.append(
tgt.Interval(phs[nucleus_index].start_time, syll.end_time,
'R'))
# Add all the intervals to the onset rhyme tier
onset_rhyme_tier.add_annotations(onset_rhyme_intervals)
# Add the onset rhyme tier to the TextGrid
tg.add_tier(onset_rhyme_tier)
# Move syll tier after the onset_rhyme_tier
tg.delete_tier(sylls_tier_name)
tg.add_tier(sylls_tier)
tgt.write_to_file(tg,
os.path.join(output_path, title + '.TextGrid'),
format='short')
def main():
ap = argparse.ArgumentParser()
ap.add_argument(
'shift',
help='offset by which to shift the boundaries (positive or negative)',
type=float)
ap.add_argument(
'file',
help='the textgrid file',
type=str)
ap.add_argument(
'-e', '--encoding',
help='file encoding (default "utf-8")',
default='utf-8',
type=str)
ap.add_argument(
'-f', '--format',
help='the output format (default "short")',
default='short',
type=str)
ap.add_argument(
'-o', '--outfile',
help='the output file (defaults to inputfile.shifted.Extension)',
type=str)
arguments = ap.parse_args()
# Read file
try:
tg = tgt.read_textgrid(
filename=arguments.file,
encoding=arguments.encoding)
except IOError:
print('An error occurred reading file {file}'.
format(file=arguments.file))
sys.exit(1)
# Create new textgrid
if arguments.outfile is None:
basename, extension = os.path.splitext(arguments.file)
output_filename = basename + '.shifted' + extension
else:
output_filename = arguments.outfile
tg_shifted = tgt.TextGrid(filename=output_filename)
# Shift boundaries
for tier in tg:
ts = tgt.util.shift_boundaries(tier, arguments.shift, 0)
tg_shifted.add_tier(ts)
# Write file
tgt.write_to_file(
textgrid=tg_shifted,
filename=tg_shifted.filename,
format=arguments.format,
encoding=arguments.encoding)
def main():
ap = argparse.ArgumentParser()
ap.add_argument(
'shift',
help='offset by which to shift the boundaries (positive or negative)',
type=float)
ap.add_argument('file', help='the textgrid file', type=str)
ap.add_argument('-e',
'--encoding',
help='file encoding (default "utf-8")',
default='utf-8',
type=str)
ap.add_argument('-f',
'--format',
help='the output format (default "short")',
default='short',
type=str)
ap.add_argument(
'-o',
'--outfile',
help='the output file (defaults to inputfile.shifted.Extension)',
type=str)
arguments = ap.parse_args()
# Read file
try:
tg = tgt.read_textgrid(filename=arguments.file,
encoding=arguments.encoding)
except IOError:
print('An error occurred reading file {file}'.format(
file=arguments.file))
sys.exit(1)
# Create new textgrid
if arguments.outfile is None:
basename, extension = os.path.splitext(arguments.file)
output_filename = basename + '.shifted' + extension
else:
output_filename = arguments.outfile
tg_shifted = tgt.TextGrid(filename=output_filename)
# Shift boundaries
for tier in tg:
ts = tgt.util.shift_boundaries(tier, arguments.shift, 0)
tg_shifted.add_tier(ts)
# Write file
tgt.write_to_file(textgrid=tg_shifted,
filename=tg_shifted.filename,
format=arguments.format,
encoding=arguments.encoding)
def main():
# Parse the command-line arguments.
args = parse_arguments()
tg_path = args['tg_path']
offset_start = args['offset_start']
offset_end = args['offset_end']
outpath = args['outpath']
# Read the TextGrid
tg = tgt.read_textgrid(tg_path)
tg_part = tgt.TextGrid()
if offset_start is None and offset_end is None:
raise Exception('At least one of offset_start and offset_end must be specified.')
elif offset_start is None:
offset_start = tg.start_time
elif offset_end is None:
offset_end = tg.end_time
for tr in tg:
intr_part = tr.get_annotations_between_timepoints(
offset_start, offset_end)
tier_part = tgt.IntervalTier(
name=tr.name,
start_time=tr.start_time,
end_time=tr.end_time,
objects=intr_part)
tg_part.add_tier(tier_part)
if outpath is None:
tg_dirname, tg_filename = os.path.split(tg_path)
outpath = os.path.splitext(tg_filename)[0] + '_part.TextGrid'
tgt.write_to_file(tg_part, outpath)
def write_grids(self, output_path, columns, one_grid_per_match, sound_path,
left_padding, right_padding, remember_time, file_prefix):
if "coquery_invisible_origin_id" not in self.df.columns:
one_grid_per_match = True
self.output_path = output_path
grids = self.fill_grids(columns, one_grid_per_match, sound_path,
left_padding, right_padding, remember_time)
textgrids = collections.defaultdict(list)
self.n = 0
for x in grids:
grid = grids[x]
for i, tier in enumerate(grid.tiers):
try:
hashed, tab, feature = self.resource.split_resource_feature(
tier.name)
if hashed is not None:
link, res = get_by_hash(hashed)
label = getattr(res, "{}_{}".format(tab, feature))
tier.name = "{}.{}".format(res.name, label)
else:
# try to retrieve a resource label for the tier name:
tier.name = getattr(self.resource, tier.name)
except (ValueError, AttributeError):
# Failed to retrieve the tier name. This may happen if
# it's not a resource feature name, but for example an
# entry from the Query output branch.
pass
if one_grid_per_match:
match_fn, match_id = x
basename, _ = os.path.splitext(os.path.basename(match_fn))
filename = "{}_id{}".format(basename, match_id)
else:
match_fn, = x
basename, _ = os.path.splitext(os.path.basename(match_fn))
filename = basename
target = os.path.join(
output_path, "{}{}.TextGrid".format(file_prefix, filename))
tgt.write_to_file(grid, target)
self.n += 1
textgrids[basename].append((grid, filename, self._offsets[x]))
if sound_path:
import wave
from .sound import extract_sound
# FIXME:
# there should be a resource method that matches sound file names
# and text grid names.
for root, _, files in os.walk(sound_path):
for file_name in files:
basename, _ = os.path.splitext(file_name)
if basename in textgrids:
source = os.path.join(root, file_name)
for grid, grid_name, offset in textgrids[basename]:
target = os.path.join(
output_path,
"{}{}.wav".format(file_prefix, grid_name))
start = max(0, offset - left_padding)
end = offset - left_padding + grid.end_time
try:
extract_sound(source, target, start, end)
except wave.Error:
pass
import tgt grid = tgt.TextGrid(filename="test") tier = tgt.IntervalTier(start_time=0, end_time=5, name="mot") label = tgt.core.Interval(2, 3, "word") tier.add_annotation(label) grid.add_tier(tier) tgt.write_to_file(grid, "/home/leferrae/Desktop/These/test.textgrid")
def add_silences(title, input_path, output_path):
tg = tgt.read_textgrid(os.path.join(input_path, title + '.TextGrid'))
# Load name of all tiers
tier_names = tg.get_tier_names()
# Select a tier whose name contains 'phones'
phones_tier_name = [name for name in tier_names if 'phones' in name][0]
phones_tier = tg.get_tier_by_name(phones_tier_name)
# Replace all sil and sp intervals with <sil> tag
#store these intervals to a list so that we can add them to the other tiers
sil_intervals = []
for interval in phones_tier:
if interval.text == 'sil' or interval.text == 'sp':
interval.text = '<sil>'
sil_intervals.append(interval)
# WORDS TIER
# Select a tier whose name contains 'words'
words_tier_name = [name for name in tier_names if 'words' in name][0]
words_tier = tg.get_tier_by_name(words_tier_name)
# Add <sil> to words tier
words_tier.add_annotations(sil_intervals)
# LEMMAS TIER
# Select a tier whose name contains 'lemmas'
lemmas_tier_name = [name for name in tier_names if 'lemmas' in name][0]
lemmas_tier = tg.get_tier_by_name(lemmas_tier_name)
# Add <sil> to lemmas tier
lemmas_tier.add_annotations(sil_intervals)
# SYLLABLES TIER
# Select a tier whose name contains 'sylls'
sylls_tier_name = [name for name in tier_names if 'sylls' in name][0]
sylls_tier = tg.get_tier_by_name(sylls_tier_name)
# Add <sil> to syllables tier
sylls_tier.add_annotations(sil_intervals)
# POS TIER
# Select a tier whose name contains 'pos'
pos_tier_name = [name for name in tier_names if 'pos' in name][0]
pos_tier = tg.get_tier_by_name(pos_tier_name)
# Add <sil> to pos tier
pos_tier.add_annotations(sil_intervals)
# OR TIER
# Select a tier whose name contains 'OR'
onset_rhyme_name = [name for name in tier_names if 'OR' in name][0]
onset_rhyme_tier = tg.get_tier_by_name(onset_rhyme_name)
# Add <sil> to OR tier
onset_rhyme_tier.add_annotations(sil_intervals)
# BP TIER
# Select a tier whose name contains 'bp'
bp_tier_name = [name for name in tier_names if 'bp' in name][0]
bp_tier = tg.get_tier_by_name(bp_tier_name)
# Add <sil> to bp tier
bp_tier.add_annotations(sil_intervals)
# FP TIER
# Select a tier whose name contains 'fp'
fp_tier_name = [name for name in tier_names if 'fp' in name][0]
fp_tier = tg.get_tier_by_name(fp_tier_name)
# Add <sil> to fp tier
fp_tier.add_annotations(sil_intervals)
tgt.write_to_file(tg,
os.path.join(output_path, title + '.TextGrid'),
format='short')
print(sys.exc_info()[0])
raise
for tier in textgrid.tiers:
if tier.tier_type() == 'IntervalTier':
for i, interval in enumerate(tier.intervals):
try:
interval.end_time = t_out[int(interval.end_time * 1000)]
except:
print('Setting interval.end_time to t_out')
interval.end_time = t2[-1]
interval.start_time = t_out[int(interval.start_time * 1000)]
else:
for point in tier.points:
point.time = t_out[int(point.time * 1000)]
try:
tier.end_time = t_out[int(tier.end_time * 1000)]
except:
print('Setting tier.end_time to t_out')
tier.end_time = t2[-1]
tgt.write_to_file(textgrid,
data_folder + textgrid_dest,
format='long',
encoding='utf-16')
#%% wrap up
end_time = datetime.now()
dif_time = end_time - start_time
print('Finished all in {}'.format(dif_time))
def createTargetGrid(textGrid):
targetInterval = findMatchInterval(textGrid, findTarget(textGrid))
st = targetInterval.start_time
et = targetInterval.end_time
targetTier = tgt.IntervalTier(start_time=st,
end_time=et,
name="Target Word")
targetTier.add_interval(targetInterval)
textGrid.add_tier(targetTier)
return textGrid
allAnnotatedGrid = [
f for f in listdir(annotatedDir) if re.search(r'TextGrid', f)
]
allAlignedGrid = [f for f in listdir(alignedDir) if re.search(r'TextGrid', f)]
list.sort(allAnnotatedGrid)
list.sort(allAlignedGrid)
for i in range(len(allAnnotatedGrid)):
annotatedTextGrid = tgt.read_textgrid(annotatedDir + allAnnotatedGrid[i])
alignedTextGrid = tgt.read_textgrid(alignedDir + allAlignedGrid[i])
outGrid = annotate(alignedTextGrid, annotatedTextGrid)
outGrid = createTargetGrid(outGrid)
print(allAlignedGrid[i])
tgt.write_to_file(outGrid,
outDir + allAlignedGrid[i] + '_merged' + '.TextGrid',
format='short')
def annotate(title, xml_path, textgrid_path, annotations_path):
try:
tree = ET.parse(os.path.join(xml_path, title + '.xml'))
root = tree.getroot()
stress_phone_seq = [
] # The content here come from the xml file. Output format: [[ph1, ph2, etc.], [ph1, ph2, etc.], etc]
stress_seq = [
] # The content here come from the xml file. Output format: [[0], [2], [1],etc]
for p in root[0]:
for s in p:
for phrase in s:
for word in phrase:
# get rid of words in xml that lack a phonemic counterpart in the textGrid
if word.text not in ('!', ',', '-', '.', '..', '...',
':', '?'):
for syllable in word:
stress_phone_group = []
stress_group = []
stress_group.append(syllable.attrib['stress'])
stress_seq.append(stress_group)
for ph in syllable:
stress_phone_group.append(ph.attrib['p'])
stress_phone_seq.append(stress_phone_group)
tg = tgt.read_textgrid(os.path.join(textgrid_path,
title + '.TextGrid'))
phones_tier = tg.get_tier_by_name('phones')
word_tier = tg.get_tier_by_name('words')
#word_durations = [w for w in word_tier._get_annotations() if w.text != '-'] # use this instead of the next snippet if you remove '-' from the vocabulary. Atm '-' is mapped to 'min@s'
word_durations = []
dash_intervals = []
for w in word_tier._get_annotations():
if w.text == '-':
dash_intervals.append(w)
else:
word_durations.append(w)
for dash in dash_intervals:
# Here we delete all the phone annotation that are read out as "minus", if you don't u mess up the alignment
phones_tier.delete_annotations_between_timepoints(
dash.start_time,
dash.end_time,
left_overlap=False,
right_overlap=False)
phone_durations = [
p for p in phones_tier._get_annotations() if p.text != 'sil'
]
# here we gather the phone durations following the same format as pos_phone_seq, i.e. [[ph_dur1, ph_dur2, etc.], [ph_dur1, ph_dur2, etc.], etc]
#print([j for i in stress_phone_seq for j in i])
#print([i.text for i in phone_durations])
l = []
k = -1
for i in range(0, len(stress_phone_seq)):
m = []
for j in range(0, len(stress_phone_seq[i])):
k += 1
m.append(phone_durations[k])
l.append(m)
# here we go thru this list ([[ph_dur1, ph_dur2, etc.], [ph_dur1, ph_dur2, etc.], etc]) and we keep the first and the last duration of every syllable
syl_durations = [(syl[0].start_time, syl[-1].end_time) for syl in l]
syllable_tier = tgt.IntervalTier()
syllable_tier.name = 'syllables'
syllable_tier.start_time = phones_tier.start_time
syllable_tier.end_time = phones_tier.end_time
syllable_intervals = [
tgt.Interval(syl_durations[i][0], syl_durations[i][1],
str(stress_seq[i][0]))
for i in range(0, len(syl_durations))
]
syllable_tier.add_annotations(syllable_intervals)
for phone in phones_tier:
phone.text = phone.text.replace('Q', '@@').replace('ts',
't').replace(
'sp', 'sil')
vowels = [
'@', '@@', 'a', 'aa', 'ai', 'au', 'e', 'e@', 'ei', 'i', 'i@', 'ii',
'o', 'oi', 'oo', 'ou', 'u', 'u@', 'uh', 'uu'
]
for phone in phones_tier:
if phone.text in vowels:
phone_centre = phone.start_time + (phone.end_time -
phone.start_time) / 2
phone.text = phone.text + syllable_tier.get_annotations_by_time(
phone_centre)[0].text
# For now we generate the modified TextGrids in the same folder is the old ones. Later, sent the new files into a new folder
newTitle = os.path.join(annotations_path, title + '.TextGrid')
tgt.write_to_file(tg, newTitle, format='short')
except:
pass
def add_punctuation(title, textgrid_path, txt_path, output_path):
txt = ''
with open(os.path.join(txt_path, title + '.txt'), 'r',
encoding='utf-8') as f:
for l in f:
l = ' '.join(l.split())
for char in l.replace('\n', ' ').replace('\t', ' ').lower():
txt += char
word_non_words = [detect_non_words(w) for w in txt.split()]
# Exclude non-words such as -' , - etc.
txt_words = [w for w in word_non_words if w != '<punct>']
# Strip words of punctuation before and after the first/last alphanum
txt_words = [clean_word(w, title, txt) for w in txt_words]
tg = tgt.read_textgrid(os.path.join(textgrid_path, title + '.TextGrid'))
# Load name of all tiers
tier_names = tg.get_tier_names()
# Select a tier whose name contains 'words'
word_tier_name = [name for name in tier_names if 'words' in name][0]
word_tier = tg.get_tier_by_name(word_tier_name)
word_list = [w.text for w in word_tier._get_annotations()]
if len(word_list) == len(txt_words):
w_indices = []
w_indices.append(0)
start = 0
for lw in txt_words:
idx = txt.find(lw, start, len(txt))
start = idx + len(lw)
w_indices.append(idx)
w_indices.append(idx + len(lw))
w_indices.append(len(txt))
p_indices = [[w_indices[i], w_indices[i + 1]]
for i in range(0,
len(w_indices) - 1, 2)]
punctuation = [txt[i[0]:i[1]].replace(' ', '') for i in p_indices]
punctuation[0] = 'start' + punctuation[0]
punctuation[-1] = punctuation[-1] + 'end'
punctuation = [p if p != '' else '_' for p in punctuation]
bp = punctuation[0:-1]
fp = punctuation[1:]
word_durations = []
for w in word_tier._get_annotations():
word_durations.append(w)
# here we go thru this list ([[w_dur1, w_dur2, etc.], [w_dur1, w_dur2, etc.], etc]) and we keep the first and the last duration of every word
bp_tier = tgt.IntervalTier()
bp_tier.name = [name for name in tier_names
if 'words' in name][0].replace('words', 'bp')
bp_intervals = [
tgt.Interval(word_durations[i].start_time,
word_durations[i].end_time, bp[i])
for i in range(0, len(word_durations))
]
bp_tier.add_annotations(bp_intervals)
tg.add_tier(bp_tier)
fp_tier = tgt.IntervalTier()
fp_tier.name = [name for name in tier_names
if 'words' in name][0].replace('words', 'fp')
fp_intervals = [
tgt.Interval(word_durations[i].start_time,
word_durations[i].end_time, fp[i])
for i in range(0, len(word_durations))
]
fp_tier.add_annotations(fp_intervals)
tg.add_tier(fp_tier)
else:
word_durations = []
for w in word_tier._get_annotations():
word_durations.append(w)
bp = ['start'] + ['<unk>' for i in range(len(word_durations) - 1)]
fp = ['<unk>' for i in range(len(word_durations) - 1)] + ['end']
bp_tier = tgt.IntervalTier()
bp_tier.name = [name for name in tier_names
if 'words' in name][0].replace('words', 'bp')
bp_intervals = [
tgt.Interval(word_durations[i].start_time,
word_durations[i].end_time, bp[i])
for i in range(0, len(word_durations))
]
bp_tier.add_annotations(bp_intervals)
tg.add_tier(bp_tier)
fp_tier = tgt.IntervalTier()
fp_tier.name = [name for name in tier_names
if 'words' in name][0].replace('words', 'fp')
fp_intervals = [
tgt.Interval(word_durations[i].start_time,
word_durations[i].end_time, fp[i])
for i in range(0, len(word_durations))
]
fp_tier.add_annotations(fp_intervals)
tg.add_tier(fp_tier)
# For now we generate the modified TextGrids in the same folder is the old ones. Later, sent the new files into a new folder
tgt.write_to_file(tg,
os.path.join(output_path, title + '.TextGrid'),
format='short')
def write_grids(self, output_path, columns, one_grid_per_match,
sound_path, left_padding, right_padding, remember_time,
file_prefix):
if "coquery_invisible_origin_id" not in self.df.columns:
one_grid_per_match = True
self.output_path = output_path
grids = self.fill_grids(columns, one_grid_per_match, sound_path,
left_padding, right_padding, remember_time)
textgrids = collections.defaultdict(list)
self.n = 0
for x in grids:
grid = grids[x]
for i, tier in enumerate(grid.tiers):
try:
hashed, tab, feature = self.resource.split_resource_feature(tier.name)
if hashed is not None:
link, res = get_by_hash(hashed)
label = getattr(res, "{}_{}".format(tab, feature))
tier.name = "{}.{}".format(res.name, label)
else:
# try to retrieve a resource label for the tier name:
tier.name = getattr(self.resource, tier.name)
except (ValueError, AttributeError):
# Failed to retrieve the tier name. This may happen if
# it's not a resource feature name, but for example an
# entry from the Query output branch.
pass
if one_grid_per_match:
match_fn, match_id = x
basename, _ = os.path.splitext(os.path.basename(match_fn))
filename = "{}_id{}".format(basename, match_id)
else:
match_fn, = x
basename, _ = os.path.splitext(os.path.basename(match_fn))
filename = basename
target = os.path.join(output_path, "{}{}.TextGrid".format(
file_prefix, filename))
tgt.write_to_file(grid, target)
self.n += 1
textgrids[basename].append((grid, filename, self._offsets[x]))
if sound_path:
import wave
from .sound import extract_sound
# FIXME:
# there should be a resource method that matches sound file names
# and text grid names.
for root, _, files in os.walk(sound_path):
for file_name in files:
basename, _ = os.path.splitext(file_name)
if basename in textgrids:
source = os.path.join(root, file_name)
for grid, grid_name, offset in textgrids[basename]:
target = os.path.join(
output_path,
"{}{}.wav".format(
file_prefix,
grid_name))
start = max(0, offset - left_padding)
end = offset - left_padding + grid.end_time
try:
extract_sound(source, target, start, end)
except wave.Error:
pass
# Usage: python segment_laughter.py <input_audio_file> <stored_model_path> <output_folder> <save_to_textgrid>
if __name__ == '__main__':
if parse_inputs():
input_path, model_path, output_path, threshold, min_length, save_to_textgrid = parse_inputs(
)
min_length = seconds_to_frames(min_length)
laughs = laugh_segmenter.segment_laughs(input_path, model_path,
output_path, threshold,
min_length, save_to_textgrid)
print("found %d laughs." % (len(laughs)))
if not save_to_textgrid:
for laugh in laughs:
print(laugh)
else:
tg = tgt.TextGrid()
laughs_tier = tgt.IntervalTier(
name='laughter',
objects=[
tgt.Interval(l['start'], l['end'], 'laugh') for l in laughs
])
tg.add_tier(laughs_tier)
fname = os.path.splitext(os.path.basename(input_path))[0]
tgt.write_to_file(
tg, os.path.join(output_path, fname + '_laughter.TextGrid'))
print('Saved laughter segments in {}'.format(
os.path.join(output_path, fname + '_laughter.TextGrid')))
def add_pos(title, input1_path, input2_path, output_path):
# Load the textgrid
tg = tgt.read_textgrid(os.path.join(input1_path, title + '.TextGrid'))
# Load name of all tiers
tier_names = tg.get_tier_names()
# Select a tier whose name contains 'words'
words_tier_name = [name for name in tier_names if 'words' in name][0]
words_tier = tg.get_tier_by_name(words_tier_name)
# Start an empty tier for POS_tags
pos_tier = tgt.IntervalTier()
pos_tier_name = [name for name in tier_names
if 'words' in name][0].replace('words', 'pos')
pos_tier.name = pos_tier_name
# Extract words intervals
word_intervals = [w for w in words_tier._get_annotations()]
# Extract words
words = [w.text for w in words_tier._get_annotations()]
# Load text
txt = ''
with open(os.path.join(input2_path, title + '.txt'), 'r',
encoding='utf-8') as f:
for l in f:
l = ' '.join(l.split())
for char in l.replace('\n', ' ').replace('\t', ' '):
txt += char
# Try to use my own tagger from txt and see if it matches the words in the original word tier
# If they don't match just use the list of words from the tier and feed them to the tagger (this option is less accurate)
my_tags = my_tagger(txt)
if len(my_tags) == len(words):
# True for every mismatch between words in words_tier and words produced by my_tagger
mismatches = [
True for i, tag in enumerate(my_tags) if tag[0] != words[i]
]
# If everything matches up we can use my_tags, else we resort to the vanilla nltk one
if True not in mismatches:
POS_tags = my_tags
else:
POS_tags = nltk.pos_tag(words)
else:
print(title)
POS_tags = nltk.pos_tag(words)
pos_intervals = [
tgt.Interval(interval.start_time, interval.end_time, POS_tags[i][1])
for i, interval in enumerate(word_intervals)
]
pos_tier.add_annotations(pos_intervals)
tg.add_tier(pos_tier)
tgt.write_to_file(tg,
os.path.join(output_path, title + '.TextGrid'),
format='short')
def stitch_textgrid(batch_title, sequenced_title, input2b_path, input2_path,
output3_path):
combined_intervals = []
new_tg = tgt.TextGrid()
new_phone_tier = tgt.IntervalTier()
final_phone_tier = tgt.IntervalTier()
new_word_tier = tgt.IntervalTier()
last_dur = 0.0
for i, title in enumerate(sequenced_title):
wave_file = wave.open(os.path.join(input2b_path, title + '.wav'), 'rb')
frameRate = wave_file.getframerate()
n_frames = wave_file.getnframes()
dur = n_frames / frameRate
f0_start_time = 0.0
f0_end_time = dur
tg = tgt.read_textgrid(os.path.join(input2_path, title + '.TextGrid'))
# Load name of all tiers
tier_names = tg.get_tier_names()
words_tier_name = [name for name in tier_names if 'words' in name][0]
words_tier = tg.get_tier_by_name(words_tier_name)
phones_tier_name = [name for name in tier_names if 'phones' in name][0]
phones_tier = tg.get_tier_by_name(phones_tier_name)
word_annotations = words_tier.get_annotations_between_timepoints(
f0_start_time, f0_end_time)
phone_annotations = phones_tier.get_annotations_between_timepoints(
f0_start_time, f0_end_time)
word_intervals = []
for interval in word_annotations:
interval.end_time = interval.end_time + last_dur
interval.start_time = interval.start_time + last_dur
word_intervals.append(interval)
if word_intervals[-1].end_time > last_dur + f0_end_time:
word_intervals[-1].end_time = last_dur + f0_end_time
phone_intervals = []
for j, interval in enumerate(phone_annotations):
interval.end_time = interval.end_time + last_dur
interval.start_time = interval.start_time + last_dur
if interval.text != 'sil' and interval.text != 'sp':
phone_intervals.append(interval)
elif i == len(sequenced_title) - 1 and j == len(
phone_annotations) - 1:
phone_intervals.append(interval)
if phone_intervals[-1].end_time > last_dur + f0_end_time:
phone_intervals[-1].end_time = last_dur + f0_end_time
new_word_tier.add_annotations(word_intervals)
new_phone_tier.add_annotations(phone_intervals)
last_dur += dur
phones_tier_copy = new_phone_tier.get_copy_with_gaps_filled(
start_time=None, end_time=None, empty_string='')
# Replace all sil and sp intervals with <sil> tag
#store these intervals to a list so that we can add them to the other tiers
sil_intervals = []
phone_intervals = []
for interval in phones_tier_copy:
if interval.text == '':
interval.text = 'sil'
sil_intervals.append(interval)
else:
phone_intervals.append(interval)
final_phone_tier.add_annotations(phone_intervals)
final_phone_tier.add_annotations(sil_intervals)
final_phone_tier.name = phones_tier_name
new_word_tier.name = words_tier_name
new_tg.add_tier(new_word_tier)
new_tg.add_tier(final_phone_tier)
tgt.write_to_file(new_tg,
os.path.join(output3_path, batch_title + '.TextGrid'),
format='short')
def add_syllables(title, input_path, syllabification_file_path, output_path):
# Load language syllable structure for the syllabifier
with open(syllabification_file_path) as f:
language_syllables = json.load(f)
# Load the textgrid
tg = tgt.read_textgrid(os.path.join(input_path,title+'.TextGrid'))
# Load name of all tiers
tier_names = tg.get_tier_names()
# Select a tier whose name contains 'words'
words_tier_name = [name for name in tier_names if 'words' in name][0]
words_tier = tg.get_tier_by_name(words_tier_name)
# Select a tier whose name contains 'phones'
phones_tier_name = [name for name in tier_names if 'phones' in name][0]
phones_tier = tg.get_tier_by_name(phones_tier_name)
# Start an empty tier for syllables
syllable_tier = tgt.IntervalTier()
syll_tier_name = [name for name in tier_names if 'words' in name][0].replace('words', 'sylls')
syllable_tier.name = syll_tier_name
# Syllabify one word at a time
for w in words_tier._get_annotations():
# For the current word, get all of its phones
phs = phones_tier.get_annotations_between_timepoints(w.start_time, w.end_time)
for ph in phs:
if ph.text == 'spn':
ph.text = 'aa1'
# Transform the string of phones into a string of syllables
# Format: ph1 ph2 . ph3 ph4 ph5 . ph6 etc.
s = stringify(syllabify(' '.join([ph.text for ph in phs]), language_syllables))
# From string of syllables to a nested lists of phone indeces
# Format: [[ph1_idx, ph2_idx, etc.], [ph3_idx, ph4_idx, etc.], etc.]
sylls = [syll.split() for syll in s.split('.')]
i = 0
sylls_indeces = []
for j, syll in enumerate(sylls):
syll_indeces = []
for k in range(0, len(syll)):
syll_indeces.append(int(i))
i += 1
sylls_indeces.append(syll_indeces)
# Extract the relevant intervals using the indeces
sylls_intervals = [[phs[index] for index in ph_group] for ph_group in sylls_indeces]
# Extract the stress for each syllable:
# Format: [['0'], ['1'], etc.]
sylls_stresses = [[char for char in ''.join(ph_group) if char.isdigit()==True] for ph_group in sylls]
sylls_stresses = [ph_group if ph_group != [] else ['0'] for ph_group in sylls_stresses]
#print(w)
#print(sylls_indeces)
#print(sylls_stresses)
#print(sylls_intervals)
syllable_intervals = [tgt.Interval(interval[0].start_time, interval[-1].end_time, str(sylls_stresses[i][0])) for i, interval in enumerate(sylls_intervals)]
#print(syllable_intervals)
syllable_tier.add_annotations(syllable_intervals)
tg.add_tier(syllable_tier)
tgt.write_to_file(tg, os.path.join(output_path,title+'.TextGrid'), format='short')
