import glob import json import os from dataclasses import dataclass import datasets from datasets import BuilderConfig, SplitGenerator _CITATION = """\ @article{yang2018scidtb, title={Scidtb: Discourse dependency treebank for scientific abstracts}, author={Yang, An and Li, Sujian}, journal={arXiv preprint arXiv:1806.03653}, year={2018} } """ _DESCRIPTION = """Annotation corpus for discourse relations benefits NLP tasks such as machine translation and question answering. SciDTB is a domain-specific discourse treebank annotated on scientific articles. Different from widely-used RST-DT and PDTB, SciDTB uses dependency trees to represent discourse structure, which is flexible and simplified to some extent but do not sacrifice structural integrity. We discuss the labeling framework, annotation workflow and some statistics about SciDTB. Furthermore, our treebank is made as a benchmark for evaluating discourse dependency parsers, on which we provide several baselines as fundamental work.""" _URL = "https://codeload.github.com/PKU-TANGENT/SciDTB/zip/refs/heads/master" _HOMEPAGE = "https://github.com/PKU-TANGENT/SciDTB" @dataclass class SciDTBConfig(BuilderConfig): """BuilderConfig for SciDTB.""" def __init__(self, subdirectory_mapping, encoding, **kwargs): super(SciDTBConfig, self).__init__(**kwargs) self.subdirectory_mapping = subdirectory_mapping self.encoding = encoding class SciDTBDataset(datasets.GeneratorBasedBuilder): """Scientific Discourse Treebank Dataset""" BUILDER_CONFIGS = [ SciDTBConfig( name="SciDTB", version=datasets.Version("1.0.0", ""), description=_DESCRIPTION, subdirectory_mapping={ "train": "SciDTB-master/dataset/train", "dev": "SciDTB-master/dataset/dev/gold", "test": "SciDTB-master/dataset/test/gold", }, encoding="utf-8-sig", ), ] def _info(self): features = datasets.Features( { "root": datasets.Sequence( { "id": datasets.Value("int32"), "parent": datasets.Value("int32"), "text": datasets.Value("string"), "relation": datasets.Value("string"), } ), "file_name": datasets.Value("string"), } ) return datasets.DatasetInfo( description=_DESCRIPTION, features=features, homepage=_HOMEPAGE, citation=_CITATION, ) def _split_generators(self, dl_manager): data_dir = dl_manager.download_and_extract(_URL) return [ SplitGenerator( name=split, # These kwargs will be passed to _generate_examples gen_kwargs={ "dir_path": os.path.join(data_dir, subdir), }, ) for split, subdir in self.config.subdirectory_mapping.items() ] def _generate_examples(self, dir_path): _files = glob.glob(f"{dir_path}/*.dep") for file_path in _files: with open(file_path, mode="r", encoding=self.config.encoding) as f: annotations = json.load(f) annotations["file_name"] = os.path.basename(file_path) yield annotations["file_name"], annotations