|
|
import datasets |
|
|
import xml.etree.cElementTree as ET |
|
|
from glob import glob |
|
|
import os |
|
|
|
|
|
_UFSAC_FILE = 'ufsac-public-2.1.tar.xz' |
|
|
|
|
|
class UFSAC(datasets.GeneratorBasedBuilder): |
|
|
|
|
|
BUILDER_CONFIG_CLASS = datasets.BuilderConfig |
|
|
|
|
|
def _info(self): |
|
|
feature = { |
|
|
'tokens': datasets.Sequence(datasets.Value('string')), |
|
|
'lemmas': datasets.Sequence(datasets.Value('string')), |
|
|
'pos_tags': datasets.Sequence(datasets.Value('string')), |
|
|
'target_idx': datasets.Value('int32'), |
|
|
'sense_keys': datasets.Sequence(datasets.Value('string')), |
|
|
} |
|
|
|
|
|
return datasets.DatasetInfo( |
|
|
features=datasets.Features(feature), |
|
|
description = 'UFSAC: the unified Sense Annotated Corpora and Tool' |
|
|
) |
|
|
|
|
|
def _split_generators(self, dl_manager): |
|
|
data_dir = dl_manager.download_and_extract(_UFSAC_FILE) |
|
|
return datasets.SplitGenerator(name = datasets.Split.TRAIN, gen_kwargs={'data_dir': data_dir}), |
|
|
|
|
|
def _generate_examples(self, data_dir): |
|
|
used_sents = set() |
|
|
count = 0 |
|
|
for file in glob(os.path.join(data_dir, 'ufsac-public-2.1/*.xml')): |
|
|
context = ET.iterparse(file, events=('start', 'end')) |
|
|
event, root = next(context) |
|
|
for event, element in context: |
|
|
if element.tag == 'paragraph': |
|
|
para = element |
|
|
if element.tag != 'sentence': |
|
|
continue |
|
|
if event == 'end' and element.tag == 'sentence': |
|
|
para.remove(element) |
|
|
sent = element |
|
|
words = sent.findall('word') |
|
|
tokens = [token.attrib['surface_form'] if 'surface_form' in token.attrib else '_' for token in words] |
|
|
sent_key = ''.join([token.lower() for token in tokens]) |
|
|
if sent_key in used_sents: |
|
|
continue |
|
|
used_sents.add(sent_key) |
|
|
lemmas = [token.attrib['lemma'] if 'lemma' in token.attrib else '_' for token in words] |
|
|
pos_tags = [token.attrib['pos'] if 'pos' in token.attrib else '_' for token in words] |
|
|
for index, word in enumerate(words): |
|
|
if 'wn30_key' in word.attrib: |
|
|
senses = word.attrib['wn30_key'].split(';') |
|
|
yield count, { |
|
|
'tokens': tokens, |
|
|
'lemmas': lemmas, |
|
|
'pos_tags': pos_tags, |
|
|
'target_idx': index, |
|
|
'sense_keys': senses |
|
|
} |
|
|
count+=1 |
|
|
|
