File size: 9,278 Bytes
c13b50a |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 |
import io
import conllu
import datasets
from seacrowd.utils.common_parser import load_ud_data_as_seacrowd_kb
from seacrowd.utils.configs import SEACrowdConfig
from seacrowd.utils import schemas
from seacrowd.utils.constants import DEFAULT_SEACROWD_VIEW_NAME, DEFAULT_SOURCE_VIEW_NAME, Licenses, Tasks
_DATASETNAME = "stb_ext"
_SOURCE_VIEW_NAME = DEFAULT_SOURCE_VIEW_NAME
_UNIFIED_VIEW_NAME = DEFAULT_SEACROWD_VIEW_NAME
_LANGUAGES = ["eng"]
_LOCAL = False
_CITATION = """\
@article{wang2019genesis,
title={From genesis to creole language: Transfer learning for singlish universal dependencies parsing and POS tagging},
author={Wang, Hongmin and Yang, Jie and Zhang, Yue},
journal={ACM Transactions on Asian and Low-Resource Language Information Processing (TALLIP)},
volume={19},
number={1},
pages={1--29},
year={2019},
publisher={ACM New York, NY, USA}
}
"""
_DESCRIPTION = """\
We adopt the Universal Dependencies protocol for constructing the Singlish dependency treebank, both as a new resource
for the low-resource languages and to facilitate knowledge transfer from English. Briefly, the STB-EXT dataset offers
a 3-times larger training set, while keeping the same dev and test sets from STB-ACL. We provide treebanks with both
gold-standard as well as automatically generated POS tags.
"""
_HOMEPAGE = "https://github.com/wanghm92/Sing_Par/tree/master/TALLIP19_dataset/treebank"
_LICENSE = Licenses.MIT.value
_PREFIX = "https://raw.githubusercontent.com/wanghm92/Sing_Par/master/TALLIP19_dataset/treebank/"
_URLS = {
"gold_pos": {
"train": _PREFIX + "gold_pos/train.ext.conll",
},
"en_ud_autopos": {"train": _PREFIX + "en-ud-autopos/en-ud-train.conllu.autoupos", "validation": _PREFIX + "en-ud-autopos/en-ud-dev.conllu.ann.auto.epoch24.upos", "test": _PREFIX + "en-ud-autopos/en-ud-test.conllu.ann.auto.epoch24.upos"},
"auto_pos_multiview": {
"train": _PREFIX + "auto_pos/multiview/train.autopos.multiview.conll",
"validation": _PREFIX + "auto_pos/multiview/dev.autopos.multiview.conll",
"test": _PREFIX + "auto_pos/multiview/test.autopos.multiview.conll",
},
"auto_pos_stack": {
"train": _PREFIX + "auto_pos/stack/train.autopos.stack.conll",
"validation": _PREFIX + "auto_pos/stack/dev.autopos.stack.conll",
"test": _PREFIX + "auto_pos/stack/test.autopos.stack.conll",
},
}
_POSTAGS = ["ADJ", "ADP", "ADV", "AUX", "CONJ", "DET", "INTJ", "NOUN", "NUM", "PART", "PRON", "PROPN", "PUNCT", "SCONJ", "SYM", "VERB", "X", "root"]
_SUPPORTED_TASKS = [Tasks.POS_TAGGING, Tasks.DEPENDENCY_PARSING]
_SOURCE_VERSION = "1.0.0"
_SEACROWD_VERSION = "2024.06.20"
def config_constructor(subset_id, schema, version):
return SEACrowdConfig(name=f"{_DATASETNAME}_{subset_id}_{schema}",
version=datasets.Version(version), description=_DESCRIPTION,
schema=schema, subset_id=subset_id)
class StbExtDataset(datasets.GeneratorBasedBuilder):
"""This is a seacrowd dataloader for the STB-EXT dataset, which offers a 3-times larger training set, while keeping
the same dev and test sets from STB-ACL. It provides treebanks with both gold-standard and automatically generated POS tags."""
BUILDER_CONFIGS = [
# source
config_constructor(subset_id="auto_pos_stack", schema="source", version=_SOURCE_VERSION),
config_constructor(subset_id="auto_pos_multiview", schema="source", version=_SOURCE_VERSION),
config_constructor(subset_id="en_ud_autopos", schema="source", version=_SOURCE_VERSION),
config_constructor(subset_id="gold_pos", schema="source", version=_SOURCE_VERSION),
# seq_label
config_constructor(subset_id="auto_pos_stack", schema="seacrowd_seq_label", version=_SEACROWD_VERSION),
config_constructor(subset_id="auto_pos_multiview", schema="seacrowd_seq_label", version=_SEACROWD_VERSION),
config_constructor(subset_id="en_ud_autopos", schema="seacrowd_seq_label", version=_SEACROWD_VERSION),
config_constructor(subset_id="gold_pos", schema="seacrowd_seq_label", version=_SEACROWD_VERSION),
# dependency parsing
config_constructor(subset_id="auto_pos_stack", schema="seacrowd_kb", version=_SEACROWD_VERSION),
config_constructor(subset_id="auto_pos_multiview", schema="seacrowd_kb", version=_SEACROWD_VERSION),
config_constructor(subset_id="en_ud_autopos", schema="seacrowd_kb", version=_SEACROWD_VERSION),
config_constructor(subset_id="gold_pos", schema="seacrowd_kb", version=_SEACROWD_VERSION),
]
DEFAULT_CONFIG_NAME = f"{_DATASETNAME}_gold_pos_source"
def _info(self):
if self.config.schema == "source":
features = datasets.Features(
{
# metadata
"sent_id": datasets.Value("string"),
"text": datasets.Value("string"),
"text_en": datasets.Value("string"),
# tokens
"id": [datasets.Value("string")],
"form": [datasets.Value("string")],
"lemma": [datasets.Value("string")],
"upos": [datasets.Value("string")],
"xpos": [datasets.Value("string")],
"feats": [datasets.Value("string")],
"head": [datasets.Value("string")],
"deprel": [datasets.Value("string")],
"deps": [datasets.Value("string")],
"misc": [datasets.Value("string")],
}
)
elif self.config.schema == "seacrowd_seq_label":
features = schemas.seq_label_features(label_names=_POSTAGS)
elif self.config.schema == "seacrowd_kb":
features = schemas.kb_features
else:
raise ValueError(f"Invalid config: {self.config.schema}")
return datasets.DatasetInfo(
description=_DESCRIPTION,
features=features,
homepage=_HOMEPAGE,
license=_LICENSE,
citation=_CITATION,
)
def _split_generators(self, dl_manager):
""" "return splitGenerators"""
urls = _URLS[self.config.subset_id]
downloaded_files = dl_manager.download_and_extract(urls)
splits = []
if "train" in downloaded_files:
splits.append(datasets.SplitGenerator(name=datasets.Split.TRAIN, gen_kwargs={"filepath": downloaded_files["train"]}))
if "validation" in downloaded_files:
splits.append(datasets.SplitGenerator(name=datasets.Split.VALIDATION, gen_kwargs={"filepath": downloaded_files["validation"]}))
if "test" in downloaded_files:
splits.append(datasets.SplitGenerator(name=datasets.Split.TEST, gen_kwargs={"filepath": downloaded_files["test"]}))
return splits
def _generate_examples(self, filepath):
def process_buffer(TextIO):
BOM = "\ufeff"
buffer = io.StringIO()
for line in TextIO:
line = line.replace(BOM, "") if BOM in line else line
buffer.write(line)
buffer.seek(0)
return buffer
with open(filepath, "r", encoding="utf-8") as data_file:
tokenlist = list(conllu.parse_incr(process_buffer(data_file)))
data_instances = []
for idx, sent in enumerate(tokenlist):
idx = sent.metadata["sent_id"] if "sent_id" in sent.metadata else idx
tokens = [token["form"] for token in sent]
txt = sent.metadata["text"] if "text" in sent.metadata else " ".join(tokens)
example = {
# meta
"sent_id": str(idx),
"text": txt,
"text_en": txt,
# tokens
"id": [token["id"] for token in sent],
"form": [token["form"] for token in sent],
"lemma": [token["lemma"] for token in sent],
"upos": [token["upos"] for token in sent],
"xpos": [token["xpos"] for token in sent],
"feats": [str(token["feats"]) for token in sent],
"head": [str(token["head"]) for token in sent],
"deprel": [str(token["deprel"]) for token in sent],
"deps": [str(token["deps"]) for token in sent],
"misc": [str(token["misc"]) for token in sent]
}
data_instances.append(example)
if self.config.schema == "source":
pass
if self.config.schema == "seacrowd_seq_label":
data_instances = list(
map(
lambda d: {
"id": d["sent_id"],
"tokens": d["form"],
"labels": d["upos"],
},
data_instances,
)
)
if self.config.schema == "seacrowd_kb":
data_instances = load_ud_data_as_seacrowd_kb(filepath, data_instances)
for key, exam in enumerate(data_instances):
yield key, exam
|