def run(align_func, dataset_path, output_path):
"""
Read the word pairs from a psa dataset, align them using the given func,
and write an output psa dataset.
If there is a word that cannot be converted to ASJP, an obviously wrong
alignment is output.
"""
dataset = AlignmentsDataset(dataset_path)
output = ['{} (PMI alignment)'.format(dataset.header)]
for word_a, word_b, original_al in dataset.data:
try:
asjp_a = ipa2asjp(word_a.ipa)
asjp_b = ipa2asjp(word_b.ipa)
except ValueError:
output.extend([
original_al.comment,
'\t'.join([word_a.lang, '-'] + list(word_a.ipa)),
'\t'.join([word_b.lang, '-'] + list(word_b.ipa)), ''
])
continue
for asjp_al in align_func(asjp_a, asjp_b):
ipa_corr = convert_alignment(word_a.ipa, word_b.ipa, asjp_al.corr)
output.extend([
original_al.comment,
'\t'.join([word_a.lang] + [pair[0] for pair in ipa_corr]),
'\t'.join([word_b.lang] + [pair[1] for pair in ipa_corr]), ''
])
with open_for_writing(output_path) as f:
f.write('\n'.join(output))
def write_alignments(alignments, path=None, header='OUTPUT'):
"""
Write a list of (Word, Word, Alignment) tuples to a psa file. The last
element of each tuple should be an Alignment named tuple from either this
or the align module.
If path is None or '-', use stdout.
"""
lines = [header]
for word_a, word_b, alignment in alignments:
field_size = str(max(len(word_a.lang), len(word_b.lang)))
lang_a = ('{:.<'+ field_size +'}').format(word_a.lang)
lang_b = ('{:.<'+ field_size +'}').format(word_b.lang)
align_a = [token if token else '-' for token, _ in alignment.corr]
align_b = [token if token else '-' for _, token in alignment.corr]
line_a = '\t'.join([lang_a] + align_a)
line_b = '\t'.join([lang_b] + align_b)
if hasattr(alignment, 'comment'):
comment = alignment.comment
else:
comment = str(word_a.concept)
lines.extend([comment, line_a, line_b, ''])
with open_for_writing(path) as f:
f.write('\n'.join(lines))
def write_pairs(path, word_pairs):
"""
Write a [] of pairs of IPA sequences to a file, one tab-separated pair per
line, with the IPA tokens separated by intervals.
"""
output = []
for pair in word_pairs:
output.append('\t'.join([' '.join(seq) for seq in pair]))
with open_for_writing(path) as f:
f.write('\n'.join(output))
def write_words(words, path=None, dialect='excel-tab', header=WordsDataset.DEFAULT_COLUMNS, tokenised=False): """ Write the words ([] of Word tuples) to a csv file using the given dialect. If path is None or '-', use stdout. The header arg should be a list of the headings for the language, concept, and transcription columns, respectively. If tokenised is set to True, separate the IPA tokens with spaces. """ joiner = ' ' if tokenised else '' with open_for_writing(path, newline='') as f: writer = csv.writer(f, dialect=dialect) writer.writerow(header) for word in words: writer.writerow([word.lang, word.concept, joiner.join(word.ipa)])
def run_sca(dataset_path, output_path):
"""
Read the word pairs from a psa dataset, align them using the SCA algorithm,
and write an output psa dataset.
"""
dataset = AlignmentsDataset(dataset_path, keep_digits=True)
output = ['{} (SCA alignment)'.format(dataset.header)]
for word_a, word_b, alignment in dataset.data:
sca = Pairwise(word_a.ipa,
word_b.ipa,
merge_vowels=False,
merge_geminates=False)
sca.align()
align_a, align_b, _ = sca.alignments[0]
output.extend([
alignment.comment, '\t'.join([word_a.lang] + align_a),
'\t'.join([word_b.lang] + align_b), ''
])
with open_for_writing(output_path) as f:
f.write('\n'.join(output))