SciGraph.py

# -*- coding: utf-8 -*-
"""
@Project    : indexing
@File       : SciGraph
@Email      : yanyuchen@zju.edu.cn
@Author     : Yan Yuchen
@Time       : 2023/3/9 12:53
"""
import json
import datasets
import pandas as pd
import numpy as np
from sklearn.model_selection import train_test_split


_CITATION = """\
@InProceedings{yan-EtAl:2022:Poster,
  author    = {Yuchen Yan and Chong Chen},
  title     = {SciGraph: A Knowledge Graph Constructed by Function and Topic Annotation of Scientific Papers},
  booktitle = {3rd Workshop on Extraction and Evaluation of Knowledge Entities from Scientific Documents (EEKE2022), June 20-24, 2022, Cologne, Germany and Online},
  month     = {June},
  year      = {2022},
  address   = {Beijing, China},
  url       = {https://ceur-ws.org/Vol-3210/paper16.pdf}
}
"""

_DESCRIPTION = """\
"""

_HOMEPAGE = ""

# The license information was obtained from https://github.com/boudinfl/ake-datasets as the dataset shared over here is taken from here
_LICENSE = ""

_URLS = {
    'classes': 'class.json',
    'function': 'assign.json',
    'topic': 'paper_new.json'
}

# TODO: Add link to the official dataset URLs here


# TODO: Name of the dataset usually match the script name with CamelCase instead of snake_case
class SciGraph(datasets.GeneratorBasedBuilder):
    """TODO: Short description of my dataset."""

    VERSION = datasets.Version("0.0.1")

    BUILDER_CONFIGS = [
        datasets.BuilderConfig(name="function", version=VERSION,
                               description="This part of my dataset covers extraction"),
        datasets.BuilderConfig(name="topic", version=VERSION,
                               description="This part of my dataset covers generation")
    ]
    
    DEFAULT_CONFIG_NAME = "function"

    def _info(self):
        classes = ['综述与进展', '论证与对比', '思考与探讨', '原理与计算', '技术与方法', '设计与应用']
        if self.config.name == "function":  # This is the name of the configuration selected in BUILDER_CONFIGS above
            features = datasets.Features(
                {
                    "id": datasets.Value("string"),
                    "abstract": datasets.Value("string"),
                    "label": datasets.features.ClassLabel(names=classes, num_classes=len(classes))
                }
            )
        else:
            features = datasets.Features(
                {
                    "id": datasets.Value("string"),
                    "abstract": datasets.Value("string"),
                    "keywords": datasets.features.Sequence(datasets.Value("string"))
                }
            )

        return datasets.DatasetInfo(
            # This is the description that will appear on the datasets page.
            description=_DESCRIPTION,
            # This defines the different columns of the dataset and their types
            features=features,
            homepage=_HOMEPAGE,
            # License for the dataset if available
            license=_LICENSE,
            # Citation for the dataset
            citation=_CITATION,
        )

    def _split_generators(self, dl_manager):
        data_dir = dl_manager.download_and_extract(_URLS)

        return [
            datasets.SplitGenerator(
                name=datasets.Split.TRAIN,
                # These kwargs will be passed to _generate_examples
                gen_kwargs={
                    "split": "train",
                    "classes": data_dir['classes'],
                    "function": data_dir['function'],
                    "topic": data_dir['topic']
                },
            ),
            datasets.SplitGenerator(
                name=datasets.Split.TEST,
                # These kwargs will be passed to _generate_examples
                gen_kwargs={
                    "split": "test",
                    "classes": data_dir['classes'],
                    "function": data_dir['function'],
                    "topic": data_dir['topic']
                },
            ),
            datasets.SplitGenerator(
                name=datasets.Split.VALIDATION,
                # These kwargs will be passed to _generate_examples
                gen_kwargs={
                    "split": "valid",
                    "classes": data_dir['classes'],
                    "function": data_dir['function'],
                    "topic": data_dir['topic']
                },
            )
        ]

    # method parameters are unpacked from `gen_kwargs` as given in `_split_generators`
    def _generate_examples(self, split, classes, function, topic):
        if self.config.name == 'function':
            with open(classes, 'r') as f:
                functions = list(json.load(f).keys())
            data = pd.read_json(function)
            data = data.loc[data[functions].sum(axis=1) == 1]
            data['label'] = [functions[row.tolist().index(1)] for index, row in data[functions].iterrows()]
            data = data[['_id', 'abstract', 'label']]
            
            
            train_data, valid_data = train_test_split(data, test_size=0.1, random_state=42)
            
            test_data = pd.read_json(function)
            test_data = test_data.loc[test_data[functions].sum(axis=1) == 0]
            if split == 'train':
                for idx, row in train_data.iterrows():
                    yield idx, {
                        "id": row._id,
                        "abstract": row.abstract,
                        "label": row.label
                    }
            elif split == 'valid':
                for idx, row in valid_data.iterrows():
                    yield idx, {
                        "id": row._id,
                        "abstract": row.abstract,
                        "label": row.label
                    }
            elif split == 'test':
                for idx, row in test_data.iterrows():
                    yield idx, {
                        "id": row._id,
                        "abstract": row.abstract,
                        "label": -1
                    }
                


        if self.config.name == 'topic':
            data = pd.read_json(topic)
            data = data.replace(to_replace=r'^\s*$', value=np.nan, regex=True).dropna(subset=['keywords'], axis=0)

            train_data, valid_data = train_test_split(data, test_size=0.1, random_state=42)
            test_data = pd.read_json(topic)
            if split == 'train':
                for idx, row in train_data.iterrows():
                    yield idx, {
                        "id": row._id,
                        "abstract": row.abstract,
                        "keywords": row.keywords.split('#%#')
                    }
            elif split == 'valid':
                for idx, row in valid_data.iterrows():
                    yield idx, {
                        "id": row._id,
                        "abstract": row.abstract,
                        "keywords": row.keywords.split('#%#')
                    }
            elif split == 'test':
                for idx, row in test_data.iterrows():
                    yield idx, {
                        "id": row._id,
                        "abstract": row.abstract,
                        "keywords": row.keywords.split('#%#')
                    }