conclugen.py

# coding=utf-8
# Copyright 2020 The HuggingFace Datasets Authors and the current dataset script contributor.
#
# 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.
"""ConcluGen Dataset"""


import json

import datasets

_CITATION = """\
@inproceedings{syed:2021,
  author    = {Shahbaz Syed and
               Khalid Al Khatib and
               Milad Alshomary and
               Henning Wachsmuth and
               Martin Potthast},
  editor    = {Chengqing Zong and
               Fei Xia and
               Wenjie Li and
               Roberto Navigli},
  title     = {Generating Informative Conclusions for Argumentative Texts},
  booktitle = {Findings of the Association for Computational Linguistics: {ACL/IJCNLP}
               2021, Online Event, August 1-6, 2021},
  pages     = {3482--3493},
  publisher = {Association for Computational Linguistics},
  year      = {2021},
  url       = {https://doi.org/10.18653/v1/2021.findings-acl.306},
  doi       = {10.18653/v1/2021.findings-acl.306}
}
"""


_DESCRIPTION = """\
The ConcluGen corpus is constructed for the task of argument summarization. It consists of 136,996 pairs of argumentative texts and their conclusions collected from the ChangeMyView subreddit, a web portal for argumentative discussions on controversial topics.

The corpus has three variants: aspects, topics, and targets. Each variation encodes the corresponding information via control codes. These provide additional argumentative knowledge for generating more informative conclusions. 
"""

_HOMEPAGE = "https://zenodo.org/record/4818134"

_LICENSE = "https://creativecommons.org/licenses/by/4.0/legalcode"


_REPO = "https://huggingface.co/datasets/webis/conclugen/resolve/main"

_URLS = {
   'base_train': f"{_REPO}/base_train.jsonl",
   'base_validation': f"{_REPO}/base_validation.jsonl",
   'base_test': f"{_REPO}/base_test.jsonl",
   'aspects_train': f"{_REPO}/aspects_train.jsonl",
   'aspects_validation': f"{_REPO}/aspects_validation.jsonl",
   'aspects_test': f"{_REPO}/aspects_test.jsonl",
   'targets_train': f"{_REPO}/targets_train.jsonl",
   'targets_validation': f"{_REPO}/targets_validation.jsonl",
   'targets_test': f"{_REPO}/targets_test.jsonl",
   'topic_train': f"{_REPO}/topic_train.jsonl",
   'topic_validation': f"{_REPO}/topic_validation.jsonl",
   'topic_test': f"{_REPO}/topic_test.jsonl"
}


class ConcluGen(datasets.GeneratorBasedBuilder):
    """382,545 arguments crawled from debate portals"""

    VERSION = datasets.Version("1.1.0")
    BUILDER_CONFIGS = [
        datasets.BuilderConfig(name="base", version=VERSION, description="The base version of the dataset  with no argumentative knowledge."),
        datasets.BuilderConfig(name="aspects", version=VERSION, description="Variation with argument aspects encoded."),
        datasets.BuilderConfig(name="targets", version=VERSION, description="Variation with conclusion targets encoded."),
        datasets.BuilderConfig(name="topic", version=VERSION, description="Variation with discussion topic encoded."),
    ]

    DEFAULT_CONFIG_NAME = "base" 

    def _info(self):
        features = datasets.Features(
            {
                "argument": datasets.Value("string"),
                "conclusion": datasets.Value("string"),
                "id": datasets.Value("string")
            }
        )
        return datasets.DatasetInfo(
            description=_DESCRIPTION,
            features=features,
            supervised_keys=None,
            homepage=_HOMEPAGE,
            license=_LICENSE,
            citation=_CITATION,
        )

    def _split_generators(self, dl_manager):
        """Returns SplitGenerators."""
        train_file = dl_manager.download(_URLS[self.config.name+"_train"])
        validation_file = dl_manager.download(_URLS[self.config.name+"_validation"])
        test_file = dl_manager.download(_URLS[self.config.name+"_test"])
        return [
            datasets.SplitGenerator(
                name=datasets.Split.TRAIN,
                gen_kwargs={
                    "data_file": train_file,
                },
            ),
            datasets.SplitGenerator(
                name=datasets.Split.VALIDATION,
                gen_kwargs={
                    "data_file": validation_file,
                },
            ),
            datasets.SplitGenerator(
                name=datasets.Split.TEST,
                gen_kwargs={
                    "data_file": test_file,
                },
            )
        ]

    def _generate_examples(self, data_file):
        """ Yields examples as (key, example) tuples. """
        with open(data_file, encoding="utf-8") as f:
            for row in f:
                data = json.loads(row)
                id_ = data['id']
                yield id_, {
                    "argument": data['argument'],
                    "conclusion": data["conclusion"],
                    "id": id_
                }