ldkp10k.py

import csv
import json
import os

import datasets
from typing import List, Any

# _SPLIT = ['train', 'test', 'valid']
_CITATION = """\
TBA
"""


_DESCRIPTION = """\
This new dataset is designed to solve kp NLP task and is crafted with a lot of care.
"""


_HOMEPAGE = ""

# TODO: Add the licence for the dataset here if you can find it
_LICENSE = ""

# TODO: Add link to the official dataset URLs here

_URLS = {
    "test": ["data/test.jsonl"],
    "train": "data/train",
    "valid": ["data/valid.jsonl"],
    
}


# TODO: Name of the dataset usually match the script name with CamelCase instead of snake_case
class LDKP10k(datasets.GeneratorBasedBuilder):
    """TODO: Short description of my dataset."""

    VERSION = datasets.Version("1.1.0")

    BUILDER_CONFIGS = [
        datasets.BuilderConfig(name="small", version=VERSION, description="This part of my dataset covers long document"),
        datasets.BuilderConfig(name="medium", version=VERSION, description="This part of my dataset covers abstract only"),
        datasets.BuilderConfig(name="large", version=VERSION, description="This part of my dataset covers abstract only")
       

    ]

    DEFAULT_CONFIG_NAME = "small"  

    def _info(self):
        _URLS['train']=['data/train'+"_"+self.config.name+".jsonl.zip"]
        if self.config.name=='large':
            _URLS['train']= ['data/train_large_'+str(x)+".jsonl.zip" for x in range(5)]
        
        features = datasets.Features(
            {
                "id": datasets.Value("string"),
                "sections": datasets.features.Sequence(datasets.Value("string")),
                "sec_text": datasets.features.Sequence(datasets.features.Sequence(datasets.Value("string"))),
                "extractive_keyphrases": datasets.features.Sequence(datasets.Value("string")),
                "abstractive_keyphrases": datasets.features.Sequence(datasets.Value("string")),
                "sec_bio_tags": datasets.features.Sequence(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):
        #print(os.listdir())
        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={
                    "filepaths": [os.path.join(x,filename)   for x in data_dir['train'] for filename in os.listdir(x)],
                    "split": "train",
                },
            ),
            datasets.SplitGenerator(
                name=datasets.Split.TEST,
                # These kwargs will be passed to _generate_examples
                gen_kwargs={
                    "filepaths": data_dir['test'],
                    "split": "test"
                },
            ),
            datasets.SplitGenerator(
                name=datasets.Split.VALIDATION,
                # These kwargs will be passed to _generate_examples
                gen_kwargs={
                    "filepaths": data_dir['valid'],
                    "split": "valid",
                },
            ),
        ]

    # method parameters are unpacked from `gen_kwargs` as given in `_split_generators`
    def _generate_examples(self, filepaths, split):
        for filepath in filepaths:
            with open(filepath, encoding="utf-8") as f:
                for key, row in enumerate(f):
                    data = json.loads(row)
                    yield key, {
                        "id": data['paper_id'],
                        "sections": data["sections"],
                        "sec_text": data["sec_text"],
                        "extractive_keyphrases": data["extractive_keyphrases"],
                        "abstractive_keyphrases": data["abstractive_keyphrases"],
                        "sec_bio_tags": data["sec_bio_tags"]
                    }