def load_textgrid(self, path):
tg = TextGrid()
try:
tg.read(path)
except ValueError as e:
raise (TextGridError('The file {} could not be parsed: {}'.format(
path, str(e))))
return tg
def guess_textgrid_format(path):
"""
Given a directory, tries to guess what format the textgrids are in
Parameters
----------
path : str
the path of the directory containing the textgrids
Returns
-------
str or None
textgrid format or None if file is not textgrid and directory doesn't contain textgrids
"""
from .inspect import inspect_labbcat, inspect_mfa, inspect_fave
if os.path.isdir(path):
counts = {'mfa': 0, 'labbcat': 0, 'fave': 0, None: 0}
for root, subdirs, files in os.walk(path):
for f in files:
if not f.lower().endswith('.textgrid'):
continue
tg_path = os.path.join(root, f)
tg = TextGrid()
try:
tg.read(tg_path)
except ValueError as e:
raise (TextGridError(
'The file {} could not be parsed: {}'.format(
tg_path, str(e))))
labbcat_parser = inspect_labbcat(tg_path)
mfa_parser = inspect_mfa(tg_path)
fave_parser = inspect_fave(tg_path)
if labbcat_parser._is_valid(tg):
counts['labbcat'] += 1
elif mfa_parser._is_valid(tg):
counts['mfa'] += 1
elif fave_parser._is_valid(tg):
counts['fave'] += 1
else:
counts[None] += 1
return max(counts.keys(), key=lambda x: counts[x])
elif path.lower().endswith('.textgrid'):
tg = TextGrid()
tg.read(path)
labbcat_parser = inspect_labbcat(path)
mfa_parser = inspect_mfa(path)
fave_parser = inspect_fave(path)
if labbcat_parser._is_valid(tg):
return 'labbcat'
elif mfa_parser._is_valid(tg):
return 'mfa'
elif fave_parser._is_valid(tg):
return 'fave'
return None
def parse_discourse(self, path, types_only=False):
'''
Parse a TextGrid file for later importing.
Parameters
----------
path : str
Path to TextGrid file
Returns
-------
:class:`~polyglotdb.io.discoursedata.DiscourseData`
Parsed data from the file
'''
tg = self.load_textgrid(path)
if len(tg.tiers) != len(self.annotation_types):
raise (TextGridError(
"The TextGrid ({}) does not have the same number of interval tiers as the number of annotation types specified."
.format(path)))
name = os.path.splitext(os.path.split(path)[1])[0]
if self.speaker_parser is not None:
speaker = self.speaker_parser.parse_path(path)
else:
speaker = None
for a in self.annotation_types:
a.reset()
a.speaker = speaker
#Parse the tiers
for i, ti in enumerate(tg.tiers):
if isinstance(ti, IntervalTier):
self.annotation_types[i].add(
((x.mark.strip(), x.minTime, x.maxTime) for x in ti))
else:
self.annotation_types[i].add(
((x.mark.strip(), x.time) for x in ti))
pg_annotations = self._parse_annotations(types_only)
data = DiscourseData(name, pg_annotations, self.hierarchy)
for a in self.annotation_types:
a.reset()
data.wav_path = find_wav_path(path)
return data
def parse_discourse(self, path, types_only=False):
'''
Parse a TextGrid file for later importing.
Parameters
----------
path : str
Path to TextGrid file
Returns
-------
:class:`~polyglotdb.io.discoursedata.DiscourseData`
Parsed data from the file
'''
tg = TextGrid()
tg.read(path)
if not self._is_valid(tg):
raise (
TextGridError('This file cannot be parsed by the MFA parser.'))
name = os.path.splitext(os.path.split(path)[1])[0]
if self.speaker_parser is not None:
speaker = self.speaker_parser.parse_path(path)
else:
speaker = None
for a in self.annotation_types:
a.reset()
a.speaker = speaker
#Parse the tiers
for i, ti in enumerate(tg.tiers):
if ti.name == 'words':
self.annotation_types[0].add(
((x.mark.strip(), x.minTime, x.maxTime) for x in ti))
elif ti.name == 'phones':
self.annotation_types[1].add(
((x.mark.strip(), x.minTime, x.maxTime) for x in ti))
pg_annotations = self._parse_annotations(types_only)
data = DiscourseData(name, pg_annotations, self.hierarchy)
for a in self.annotation_types:
a.reset()
data.wav_path = find_wav_path(path)
return data
def parse_discourse(self, path, types_only=False):
'''
Parse a TextGrid file for later importing.
Parameters
----------
path : str
Path to TextGrid file
Returns
-------
:class:`~polyglotdb.io.discoursedata.DiscourseData`
Parsed data from the file
'''
tg = self.load_textgrid(path)
if not self._is_valid(tg):
raise (TextGridError(
'The file "{}" cannot be parsed by the LaBB-CAT parser.'.
format(path)))
name = os.path.splitext(os.path.split(path)[1])[0]
self.speaker_parser = DirectorySpeakerParser()
if self.speaker_parser is not None:
speaker = self.speaker_parser.parse_path(path)
else:
speaker = None
for a in self.annotation_types:
a.reset()
a.speaker = speaker
#Parse the tiers
for i, ti in enumerate(tg.tiers):
if ti.name.startswith('transcrip'):
self.annotation_types[0].add(
((x.mark.strip(), x.minTime, x.maxTime) for x in ti))
elif ti.name.startswith('segment'):
self.annotation_types[1].add(
((x.mark.strip(), x.minTime, x.maxTime) for x in ti))
pg_annotations = self._parse_annotations(types_only)
data = DiscourseData(name, pg_annotations, self.hierarchy)
for a in self.annotation_types:
a.reset()
data.wav_path = find_wav_path(path)
return data
def parse_discourse(self, path, types_only=False):
'''
Parse a TextGrid file for later importing.
Parameters
----------
path : str
Path to TextGrid file
Returns
-------
:class:`~polyglotdb.io.discoursedata.DiscourseData`
Parsed data from the file
'''
tg = TextGrid()
tg.read(path)
if not self._is_valid(tg):
raise (TextGridError(
'The file "{}" cannot be parsed by the FAVE parser.'.format(
path)))
name = os.path.splitext(os.path.split(path)[1])[0]
dummy = self.annotation_types
self.annotation_types = []
#Parse the tiers
for i, ti in enumerate(tg.tiers):
try:
speaker, type = ti.name.split(' - ')
except ValueError:
continue
at = OrthographyTier(type, type)
at.speaker = speaker
at.add(((x.mark.strip(), x.minTime, x.maxTime) for x in ti))
self.annotation_types.append(at)
pg_annotations = self._parse_annotations(types_only)
data = DiscourseData(name, pg_annotations, self.hierarchy)
data.wav_path = find_wav_path(path)
self.annotation_types = dummy
return data
def load_textgrid(self, path):
"""
Load a TextGrid file
Parameters
----------
path : str
Path to the TextGrid file
Returns
-------
:class:`~textgrid.TextGrid`
TextGrid object
"""
tg = TextGrid()
try:
tg.read(path)
except ValueError as e:
raise (TextGridError('The file {} could not be parsed: {}'.format(
path, str(e))))
return tg
def parse_discourse(self, path, types_only=False):
'''
Parse a TextGrid file for later importing.
Parameters
----------
path : str
Path to TextGrid file
Returns
-------
:class:`~polyglotdb.io.discoursedata.DiscourseData`
Parsed data from the file
'''
tg = TextGrid()
try:
tg.read(path)
except Exception as e:
print('There was an issue parsing {}:'.format(path))
raise
multiple_speakers, is_valid = self._is_valid(tg)
if not is_valid:
raise (TextGridError('This file ({}) cannot be parsed by the {} parser.'.format(path, self.name)))
name = os.path.splitext(os.path.split(path)[1])[0]
# Format 1
if not multiple_speakers:
if self.speaker_parser is not None:
speaker = self.speaker_parser.parse_path(path)
else:
speaker = None
for a in self.annotation_types:
a.reset()
a.speaker = speaker
# Parse the tiers
for i, ti in enumerate(tg.tiers):
if ti.name.lower().startswith(self.word_label):
self.annotation_types[0].add(((x.mark.strip(), x.minTime, x.maxTime) for x in ti))
elif ti.name.lower().startswith(self.phone_label):
self.annotation_types[1].add(((x.mark.strip(), x.minTime, x.maxTime) for x in ti))
pg_annotations = self._parse_annotations(types_only)
data = DiscourseData(name, pg_annotations, self.hierarchy)
for a in self.annotation_types:
a.reset()
# Format 2
else:
dummy = self.annotation_types
self.annotation_types = []
wav_path = find_wav_path(path)
speaker_channel_mapping = {}
if wav_path is not None:
n_channels = get_n_channels(wav_path)
if n_channels > 1:
# Figure speaker-channel mapping
n_tiers = 0
for ti in tg.tiers:
try:
speaker, type = ti.name.split(' - ')
except ValueError:
continue
n_tiers += 1
ind = 0
cutoffs = [x / n_channels for x in range(1, n_channels)]
for ti in tg.tiers:
try:
if self.speaker_first:
speaker, type = ti.name.split(' - ')
else:
type, speaker = ti.name.split(' - ')
except ValueError:
continue
if speaker in speaker_channel_mapping:
continue
for i, c in enumerate(cutoffs):
if ind / n_channels < c:
speaker_channel_mapping[speaker] = i
break
else:
speaker_channel_mapping[speaker] = i + 1
ind += 1
# Parse the tiers
for ti in tg.tiers:
try:
if self.speaker_first:
speaker, type = ti.name.split(' - ')
else:
type, speaker = ti.name.split(' - ')
except ValueError:
continue
if type.lower().startswith(self.word_label):
type = 'word'
elif type.lower().startswith(self.phone_label):
type = 'phone'
if len(ti) == 1 and ti[0].mark.strip() == '':
continue
at = OrthographyTier(type, type)
at.speaker = speaker
at.add(((x.mark.strip(), x.minTime, x.maxTime) for x in ti))
self.annotation_types.append(at)
pg_annotations = self._parse_annotations(types_only)
data = DiscourseData(name, pg_annotations, self.hierarchy)
data.speaker_channel_mapping = speaker_channel_mapping
self.annotation_types = dummy
data.wav_path = find_wav_path(path)
return data
def parse_discourse(self, path, types_only=False):
'''
Parse a TextGrid file for later importing.
Parameters
----------
path : str
Path to TextGrid file
Returns
-------
:class:`~polyglotdb.io.discoursedata.DiscourseData`
Parsed data from the file
'''
tg = TextGrid()
tg.read(path)
if not self._is_valid(tg):
raise (TextGridError(
'The file "{}" cannot be parsed by the FAVE parser.'.format(
path)))
name = os.path.splitext(os.path.split(path)[1])[0]
dummy = self.annotation_types
self.annotation_types = []
wav_path = find_wav_path(path)
speaker_channel_mapping = {}
if wav_path is not None:
n_channels = get_n_channels(wav_path)
if n_channels > 1:
#Figure speaker-channel mapping
n_tiers = 0
for ti in tg.tiers:
try:
speaker, type = ti.name.split(' - ')
except ValueError:
continue
n_tiers += 1
ind = 0
cutoffs = [x / n_channels for x in range(1, n_channels)]
print(cutoffs)
for ti in tg.tiers:
try:
speaker, type = ti.name.split(' - ')
except ValueError:
continue
if speaker in speaker_channel_mapping:
continue
print(ind / n_channels)
for i, c in enumerate(cutoffs):
print(c)
if ind / n_channels < c:
speaker_channel_mapping[speaker] = i
break
else:
speaker_channel_mapping[speaker] = i + 1
ind += 1
#Parse the tiers
for ti in tg.tiers:
try:
speaker, type = ti.name.split(' - ')
except ValueError:
continue
if len(ti) == 1 and ti[0].mark.strip() == '':
continue
at = OrthographyTier(type, type)
at.speaker = speaker
at.add(((x.mark.strip(), x.minTime, x.maxTime) for x in ti))
self.annotation_types.append(at)
pg_annotations = self._parse_annotations(types_only)
data = DiscourseData(name, pg_annotations, self.hierarchy)
data.speaker_channel_mapping = speaker_channel_mapping
data.wav_path = wav_path
self.annotation_types = dummy
return data
def parse_discourse(self, path, types_only=False):
"""
Parse a TextGrid file for later importing.
Parameters
----------
path : str
Path to TextGrid file
types_only : bool
Flag for whether to only save type information, ignoring the token information
Returns
-------
:class:`~polyglotdb.io.discoursedata.DiscourseData`
Parsed data from the file
"""
tg = self.load_textgrid(path)
if len(tg.tiers) != len(self.annotation_tiers):
raise (TextGridError(
"The TextGrid ({}) does not have the same number of interval tiers as the number of annotation types specified."
.format(path)))
name = os.path.splitext(os.path.split(path)[1])[0]
if self.speaker_parser is not None:
speaker = self.speaker_parser.parse_path(path)
else:
speaker = None
for a in self.annotation_tiers:
a.reset()
a.speaker = speaker
# Parse the tiers
for i, ti in enumerate(tg.tiers):
if isinstance(ti, IntervalTier):
self.annotation_tiers[i].add(
((x.mark.strip(), x.minTime, x.maxTime) for x in ti))
else:
self.annotation_tiers[i].add(
((x.mark.strip(), x.time) for x in ti))
is_empty_textgrid = True
for t in self.annotation_tiers:
for interval in t:
if isinstance(interval, Orthography):
if interval.label != "":
is_empty_textgrid = False
break
if isinstance(interval, Transcription):
if interval._list != []:
is_empty_textgrid = False
break
if is_empty_textgrid:
return None
pg_annotations = self._parse_annotations(types_only)
data = DiscourseData(name, pg_annotations, self.hierarchy)
for a in self.annotation_tiers:
a.reset()
data.wav_path = find_wav_path(path)
return data
