def to_eaf(self, skipempty=True, pointlength=0.1):
"""Convert the object to an pympi.Elan.Eaf object
:param int pointlength: Length of respective interval from points in
seconds
:param bool skipempty: Skip the empty annotations
:returns: :class:`pympi.Elan.Eaf` object
:raises ImportError: If the Eaf module can't be loaded.
:raises ValueError: If the pointlength is not strictly positive.
"""
from pympi.Elan import Eaf
eaf_out = Eaf()
if pointlength <= 0:
raise ValueError('Pointlength should be strictly positive')
for tier in self.get_tiers():
eaf_out.add_tier(tier.name)
for ann in tier.get_intervals(True):
if tier.tier_type == 'TextTier':
ann = (ann[0], ann[0]+pointlength, ann[1])
if ann[2].strip() or not skipempty:
eaf_out.add_annotation(tier.name, int(round(ann[0]*1000)),
int(round(ann[1]*1000)), ann[2])
return eaf_out
def to_eaf(self, skipempty=True, pointlength=0.1):
"""Convert the object to an pympi.Elan.Eaf object
:param int pointlength: Length of respective interval from points in
seconds
:param bool skipempty: Skip the empty annotations
:returns: :class:`pympi.Elan.Eaf` object
:raises ImportError: If the Eaf module can't be loaded.
:raises ValueError: If the pointlength is not strictly positive.
"""
from pympi.Elan import Eaf
eaf_out = Eaf()
if pointlength <= 0:
raise ValueError('Pointlength should be strictly positive')
for tier in self.get_tiers():
eaf_out.add_tier(tier.name)
for ann in tier.get_intervals(True):
if tier.tier_type == 'TextTier':
ann = (ann[0], ann[0] + pointlength, ann[1])
if ann[2].strip() or not skipempty:
eaf_out.add_annotation(tier.name,
int(round(ann[0] * 1000)),
int(round(ann[1] * 1000)), ann[2])
return eaf_out
seg2tier = {}
with open(seg2tier_path) as f:
for l in f:
tok = l.strip().split()
seg2tier[tok[0]] = tok[1]
tiers = read_ctm(words_ctm_path, seg2tier, eaf)
for tier in tiers.keys():
part = eaf.tiers[tier][2]['PARTICIPANT']
t = eaf.add_tier('{}_words'.format(tier), parent='tier', part=part, ann='Clarin-PL-service')
for tier, segs in tiers.items():
for seg in segs:
eaf.add_annotation('{}_words'.format(tier), seg[0], seg[1], seg[2])
if args.phones_ctm:
phones_ctm_path = args.phones_ctm
tiers = read_ctm(phones_ctm_path, seg2tier, eaf)
for tier in tiers.keys():
part = eaf.tiers[tier][2]['PARTICIPANT']
t = eaf.add_tier('{}_phones'.format(tier), parent='tier', part=part, ann='Clarin-PL-service')
for tier, segs in tiers.items():
for seg in segs:
eaf.add_annotation('{}_phones'.format(tier), seg[0], seg[1], seg[2])
to_eaf(str(eaf_out_path), eaf)
def make_elans(input_dir: str, output_dir: str, copy_wavs: bool):
"""
Make ELAN files based on filenames of WAV files and annotation from matching text file
:param input_dir: Directory name of folder containing TXT and WAV audio files
:param output_dir: Directory name to save EAF files into
:param copy_wavs: Setting whether or not to copy the WAV file to the output dir
"""
# Process each file
files = glob.glob(f'{input_dir}/**/*.txt', recursive=True)
print(files)
for filename in files:
filepath, ext = os.path.splitext(filename)
basename = os.path.splitext(os.path.basename(filepath))[0]
subdirname = os.path.basename(os.path.dirname(filepath))
sex = subdirname[0]
participant = subdirname[1:]
# SEX :== m | f
# SPEAKER_ID :== <INITIALS><DIGIT>
# INITIALS :== speaker initials, 3 letters
# DIGIT :== number 0-9 to differentiate speakers with identical initials
# print(filename) # input/dr1/fmem0/sa2.txt
# print(filepath) # input/dr1/fmem0/sa2
# print(subdirname) # fmem0
# print(basename) # sa2
# print(ext) # txt
# Get audio file duration - use this as the EAF annotation's end timeslot
# duration = int(librosa.get_duration(filename=os.path.join(input_dir, filename))*1000)
# Get annotation from the text file matching on file basename
with open(filename, 'r', encoding='utf-8') as text_file:
annotation = text_file.read()
annotation_split = annotation.split()
start = int(annotation_split[0])
duration = int(annotation_split[1])
# convert audio samples to seconds to ms
duration = int(duration / 16000 * 1000)
annotation_text = " ".join(annotation_split[2:])
# Add any annotation cleaning here
# annotation = re.sub(r"(\d+)", lambda x: num2words.num2words(int(x.group(0))), annotation)
print(start, duration, annotation_text)
# Make EAF file
output_eaf = Eaf()
output_eaf.add_tier('default', part=participant)
output_eaf.add_annotation('default', start, duration, annotation_text)
output_eaf.add_linked_file(
os.path.join(output_dir, f'{subdirname}-{basename}.wav'))
output_eaf.to_file(
os.path.join(output_dir, f'{subdirname}-{basename}.eaf'))
# Copy WAV?
# if copy_wavs:
shutil.copyfile(
f'{filepath}.wav',
os.path.join(output_dir, f'{subdirname}-{basename}.wav'))
print('>>> Done')
