housing_qa.py

import csv

import datasets
from datasets import Dataset, DatasetInfo, Features, ClassLabel, Value, Sequence, DatasetDict
import json
from io import StringIO
from pathlib import Path
import pandas as pd

_CITATION = """"""
_DESCRIPTION = """"""
_HOMEPAGE = ""
_URLS = {
    "questions": "data/questions.json.zip",
    "questions_aux": "data/questions_aux.json.zip",
    "statutes": "data/statutes.tsv.zip",
}

_CONFIGS = {}

_CONFIGS["questions"] = {
    "description": "Questions about housing law.",
    "features" : Features({
        'idx': Value('int32'),
        'state': Value('string'),
        'question': Value('string'),
        'answer': Value('string'),
        'question_group': Value('int32'),
        'statutes': [{
            'statute_idx': Value('int32'),
            'citation': Value('string'),
            'excerpt': Value('string'),
        }],
        'original_question': Value('string'),
        'caveats': Sequence(Value('string')),
    }),
    "license": None,
}

_CONFIGS["questions_aux"] = {
    "description": "An auxilliary set of larger questions about housing law, without statutory annotations.",
    "features" : Features({
        'idx': Value('int32'),
        'state': Value('string'),
        'question': Value('string'),
        'answer': Value('string'),
        'question_group': Value('int32'),
        'statutes': Sequence({
            'citation': Value('string'),
            'excerpt': Value('string'),
        }),
        'original_question': Value('string'),
        'caveats': Sequence(Value('string')),
    }),
    "license": None,
}

_CONFIGS["statutes"] = {
    "description": "Corpus of statutes",
    "features": Features({
        "citation": datasets.Value("string"),
        "path": datasets.Value("string"),
        "state": datasets.Value("string"),
        "text": datasets.Value("string"),
        "idx": datasets.Value("int32"),
    }),
    "license": None,
}


class HousingQA(datasets.GeneratorBasedBuilder):
    """TODO"""

    BUILDER_CONFIGS = [
        datasets.BuilderConfig(
            name=task, version=datasets.Version("1.0.0"), description=task,
        )
        for task in _CONFIGS
    ]

    def _info(self):
        features = _CONFIGS[self.config.name]["features"]
        return datasets.DatasetInfo(
            description=_DESCRIPTION,
            features=features,
            homepage=_HOMEPAGE,
            citation=_CITATION,
            license=_CONFIGS[self.config.name]["license"],
        )

    def _split_generators(self, dl_manager):
        """Returns SplitGenerators."""
        downloaded_file_dir = Path(dl_manager.download_and_extract(_URLS[self.config.name]))
        return [
            datasets.SplitGenerator(
                name="corpus" if self.config.name == "statutes" else "test",
                gen_kwargs={
                    "downloaded_file_dir": downloaded_file_dir,
                    "name": self.config.name,
                },
            ),
        ]

    def _generate_examples(self, downloaded_file_dir, name):
        """Yields examples as (key, example) tuples."""
        
        if name in ["questions", "questions_aux"]:
            fpath = downloaded_file_dir / f"{name}.json"
            data = json.loads(fpath.read_text())
            for id_line, data in enumerate(data):
                yield id_line, data

        if name in ["statutes"]:
            fpath = downloaded_file_dir / f"{name}.tsv"
            data = pd.read_csv(fpath, sep="\t", dtype={'index': 'int32'})
            data = data.to_dict(orient="records")
            for id_line, data in enumerate(data):
                yield id_line, data