# Copyright 2022 san kim # # Licensed under the Apache License, Version 2.0 (the "License"); # you may not use this file except in compliance with the License. # You may obtain a copy of the License at # # http://www.apache.org/licenses/LICENSE-2.0 # # Unless required by applicable law or agreed to in writing, software # distributed under the License is distributed on an "AS IS" BASIS, # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. # See the License for the specific language governing permissions and # limitations under the License. import os import glob import json import textwrap import zipfile import unicodedata import datasets _VERSION = datasets.Version("1.0.0", "") _URL = "https://corpus.korean.go.kr/main.do" _CITATION = """\ There is no citation information """ _DESCRIPTION = """\ # 문서 요약 말뭉치 ## 소개 (버전 1.0) 문서에서 추출한 주제문과 문서를 요약한 글로 구성된 말뭉치입니다. ## Usage ```python from datasets import load_dataset raw_datasets = load_dataset( "nikl_summarization.py", "base", cache_dir="huggingface_datasets", data_dir="data", ignore_verifications=True, ) dataset_train = raw_datasets["train"] for item in dataset_train: print(item) exit() ``` ## Documentation [Link](https://rlkujwkk7.toastcdn.net/6/NIKL_SUMMARIZATION(v1.0).pdf) """ SUMMARIZATION_FNAME_LIST = [ "NIKL_SUMMARIZATION(v1.0).zip" ] NEWSPAPER_FNAME_LIST = [ "NIKL_NEWSPAPER_v2.0.zip", "NIKL_NEWSPAPER(v1.0).zip", ] def find_file_name(root_dir, fpath_list): for fpath in fpath_list: rel_path = os.path.join(root_dir, fpath) if os.path.isfile(rel_path): return rel_path return None def _sentences2sentence(sentences): return ' '.join([x.strip() for x in sentences]) def _find_fname_from_doc_dict(doc_id, doc_dict): doc_key = doc_id.split('.')[0] return doc_dict.get(doc_key, None) def _is_punctuation(char): cat = unicodedata.category(char) if cat.startswith("P"): return True return False def _page_proc(obj): raw_example = [] for paragraph in obj['paragraph']: form = paragraph['form'].strip() if len(form) > 0: if not _is_punctuation(form[-1]): form += '.' raw_example.append(form) return ' '.join(raw_example) def _find_id_from_doc_summarization(doc_id, f): doc_json = json.loads(f.read()) for doc in doc_json['document']: if doc['id'] == doc_id: return _page_proc(doc) return None def generator(fpath, doc_fpath, is_single_sent=False): with zipfile.ZipFile(doc_fpath, "r") as doc_fp: doc_dict = doc_fp.namelist() doc_dict = {os.path.splitext(os.path.basename(k))[0]:k for k in filter(lambda x: x.endswith(".json"), doc_dict)} with zipfile.ZipFile(fpath, "r") as fp: file_list = fp.namelist() file_list = filter(lambda x: x.endswith(".json"), file_list) for fname in file_list: data = json.load(fp.open(fname, "r"))["data"] for obj in data: doc_id = obj['document_id'] subclass = obj['subclass'] head = obj['head'] subhead = obj['subhead'] doc_fname = _find_fname_from_doc_dict(doc_id, doc_dict) para = _find_id_from_doc_summarization(doc_id, doc_fp.open(doc_fname, "r")) if para is not None: if is_single_sent: for idx, summary_sentences, topic_sentences in zip(range(len(obj["summary_sentences"])), obj["summary_sentences"], obj["topic_sentences"]): summary_sentences = _sentences2sentence(obj["summary_sentences"]) topic_sentences = _sentences2sentence(obj["topic_sentences"]) yield { "document_id": doc_id+str(idx), "subclass": subclass, "head": head, "subhead": subhead, "article": para, "summary_sentences": summary_sentences, "topic_sentences": topic_sentences, } else: summary_sentences = _sentences2sentence(obj["summary_sentences"]) topic_sentences = _sentences2sentence(obj["topic_sentences"]) yield { "document_id": doc_id, "subclass": subclass, "head": head, "subhead": subhead, "article": para, "summary_sentences": summary_sentences, "topic_sentences": topic_sentences, } class NIKLSummarizationConfig(datasets.BuilderConfig): """BuilderConfig for NIKLSummarizationConfig.""" def __init__(self, **kwargs): """BuilderConfig for NIKLSummarizationConfig. Args: **kwargs: keyword arguments forwarded to super. """ super(NIKLSummarizationConfig, self).__init__(**kwargs) class NIKLSummarization(datasets.GeneratorBasedBuilder): """NIKLSummarization Dataset""" BUILDER_CONFIGS = [ NIKLSummarizationConfig( name="base", version=datasets.Version("1.0.0"), description="NIKL Summarization dataset, concat 3 lines", ), NIKLSummarizationConfig( name="single", version=datasets.Version("1.0.0"), description="NIKL Summarization dataset, single sentence", ), ] DEFAULT_CONFIG_NAME = "base" manual_download_instructions = textwrap.dedent(f""" You need to manually download the data file on NIKL (국립국어원 모두의 말뭉치) (${_URL}). The folder containing the saved file can be used to load the dataset via 'datasets.load_dataset("nikl_summarization.py", data_dir="")' """) def _info(self): return datasets.DatasetInfo( description=_DESCRIPTION, features=datasets.Features( { "document_id": datasets.Value("string"), "subclass": datasets.Value("string"), "head": datasets.Value("string"), "subhead": datasets.Value("string"), "article": datasets.Value("string"), "summary_sentences": datasets.Value("string"), "topic_sentences": datasets.Value("string"), } ), supervised_keys=None, # Probably needs to be fixed. homepage=_URL, citation=_CITATION, ) def _split_generators(self, dl_manager: datasets.DownloadManager): summarization_fpath = find_file_name(dl_manager.manual_dir, SUMMARIZATION_FNAME_LIST) newspaper_fpath = find_file_name(dl_manager.manual_dir, NEWSPAPER_FNAME_LIST) if summarization_fpath is None: raise ValueError(f"Can't find summarization file({SUMMARIZATION_FNAME_LIST}) in {dl_manager.manual_dir}.") elif newspaper_fpath is None: raise ValueError(f"Can't find newspaper files({NEWSPAPER_FNAME_LIST}) in {dl_manager.manual_dir}.") path_kv = { datasets.Split.TRAIN: (summarization_fpath, newspaper_fpath), } return [ datasets.SplitGenerator(name=k, gen_kwargs={'fpath': v1, 'doc_fpath': v2}) for k, (v1, v2) in path_kv.items() ] def _generate_examples(self, fpath, doc_fpath): """Yields examples.""" is_single_sent = True if "single" in self.config.name else False for idx, item in enumerate(generator(fpath, doc_fpath, is_single_sent)): yield idx, item