def load_fragments_within(fname, verbose):
with verb_print(' loading folds within', verbose, True, True, True):
fragments = load_split(fname, multiple=True)
return fragments
def load_disc(fname, corpus, split_file, truncate, verbose):
with verb_print(' loading discovered classes', verbose, True, True, True):
split_mapping = load_split(split_file)
disc, errors = _load_classes(fname, corpus, split_mapping)
if not truncate:
errors_found = len(errors) > 0
if len(errors) > 100:
print 'There were more than 100 interval errors found.'
print 'Printing only the first 100.'
print
errors = errors[:100]
for fragment in sorted(errors,
key=lambda x: (x.name, x.interval.start)):
print ' error: {0} [{1:.3f}, {2:.3f}]'.format(
fragment.name, fragment.interval.start,
fragment.interval.end)
if not truncate and errors_found:
print 'There were errors in {0}. Use option -f to'\
' automatically skip invalid intervals.'.format(fname)
sys.exit()
if truncate:
with verb_print(' checking discovered classes and truncating'):
disc, filename_errors, interval_errors = \
truncate_intervals(disc, corpus,
split_mapping)
else:
with verb_print(' checking discovered classes', verbose, True, True,
True):
filename_errors, interval_errors = \
check_intervals(disc, split_mapping)
if not truncate:
filename_errors = sorted(filename_errors,
key=lambda x: (x.name, x.interval.start))
interval_errors = sorted(interval_errors,
key=lambda x: (x.name, x.interval.start))
interval_error = len(interval_errors) > 0
filename_error = len(filename_errors) > 0
errors_found = filename_error or interval_error
if interval_error:
print banner('intervals found in {0} outside of valid'
' splits'.format(fname))
if len(interval_errors) > 100:
print 'There were more than 100 interval errors found.'
print 'Printing only the first 100.'
print
interval_errors = interval_errors[:100]
for fragment in sorted(interval_errors,
key=lambda x: (x.name, x.interval.start)):
print ' error: {0} [{1:.3f}, {2:.3f}]'.format(
fragment.name, fragment.interval.start,
fragment.interval.end)
if filename_error:
print banner('unknown filenames found in {0}'.format(fname))
if len(filename_errors) > 100:
print 'There were more than 100 filename errors found.'
print 'Printing only the first 100.'
print
filename_errors = filename_errors[:100]
for fragment in sorted(filename_errors,
key=lambda x: (x.name, x.interval.start)):
print ' error: {0}'.format(fragment.name)
if not truncate and errors_found:
print 'There were errors in {0}. Use option -f to automatically skip invalid intervals.'.format(
fname)
sys.exit()
return disc
def load_disc(fname, corpus, split_file, truncate, verbose):
with verb_print(' loading discovered classes',
verbose, True, True, True):
split_mapping = load_split(split_file)
disc, errors = _load_classes(fname, corpus, split_mapping)
if not truncate:
errors_found = len(errors) > 0
if len(errors) > 100:
print 'There were more than 100 interval errors found.'
print 'Printing only the first 100.'
print
errors = errors[:100]
for fragment in sorted(errors, key=lambda x: (x.name, x.interval.start)):
print ' error: {0} [{1:.3f}, {2:.3f}]'.format(
fragment.name, fragment.interval.start, fragment.interval.end)
if not truncate and errors_found:
print 'There were errors in {0}. Use option -f to'\
' automatically skip invalid intervals.'.format(fname)
sys.exit()
if truncate:
with verb_print(' checking discovered classes and truncating'):
disc, filename_errors, interval_errors = \
truncate_intervals(disc, corpus,
split_mapping)
else:
with verb_print(' checking discovered classes', verbose, True,
True, True):
filename_errors, interval_errors = \
check_intervals(disc, split_mapping)
if not truncate:
filename_errors = sorted(filename_errors,
key=lambda x: (x.name, x.interval.start))
interval_errors = sorted(interval_errors,
key=lambda x: (x.name, x.interval.start))
interval_error = len(interval_errors) > 0
filename_error = len(filename_errors) > 0
errors_found = filename_error or interval_error
if interval_error:
print banner('intervals found in {0} outside of valid'
' splits'.format(fname))
if len(interval_errors) > 100:
print 'There were more than 100 interval errors found.'
print 'Printing only the first 100.'
print
interval_errors = interval_errors[:100]
for fragment in sorted(interval_errors,
key=lambda x: (x.name, x.interval.start)):
print ' error: {0} [{1:.3f}, {2:.3f}]'.format(
fragment.name,
fragment.interval.start, fragment.interval.end)
if filename_error:
print banner('unknown filenames found in {0}'
.format(fname))
if len(filename_errors) > 100:
print 'There were more than 100 filename errors found.'
print 'Printing only the first 100.'
print
filename_errors = filename_errors[:100]
for fragment in sorted(filename_errors,
key=lambda x: (x.name, x.interval.start)):
print ' error: {0}'.format(fragment.name)
if not truncate and errors_found:
print 'There were errors in {0}. Use option -f to automatically skip invalid intervals.'.format(fname)
sys.exit()
return disc
def load_fragments_within(fname, verbose):
with verb_print(' loading folds within',
verbose, True, True, True):
fragments = load_split(fname, multiple=True)
return fragments
# load gold phones and gold words
with verb_print(' loading word corpus file',
verbose, True, True, True):
wrd_corpus = load_corpus_txt(wrd_corpus_file)
with verb_print(' loading phone corpus file',
verbose, True, True, True):
phn_corpus = load_corpus_txt(phn_corpus_file)
# load across and withing folds
with verb_print(' loading folds cross',
verbose, True, True, True):
#fragments_cross = load_split(folds_cross_file,
# multiple=False)
intervals_vad = [load_split(vad_file,
multiple=False)]
# get list of file names from vad:
# names = load_names(vad_file)
try:
os.makedirs(dest)
except OSError:
pass
# Before loading the class file, check its consistency.
# If intervals that overlaps a little with silence are found,
# they are trimmed. If intervals that contain silences are found,
# an error is thrown and the evaluation breaks.
sil_tree = create_silence_tree(vad_file)
parse_class_file(disc_clsfile, sil_tree, dest, verbose)
# load discovered intervals and gold intervals
