fiqa_pt.py

import json
import csv
import os
import datasets

logger = datasets.logging.get_logger(__name__)

_DESCRIPTION = "FIQA Dataset"
_SPLITS = ["corpus", "queries"]

URL = ""
_URLs = {subset: URL + f"{subset}.jsonl.gz" for subset in _SPLITS}
_URLs.update({"qrel": URL + "qrels.tsv"})

class BEIR(datasets.GeneratorBasedBuilder):
    """BEIR BenchmarkDataset."""

    BUILDER_CONFIGS = [
        datasets.BuilderConfig(
            name=name,
            description=f"This is the {name} in the FiQA dataset.",
        ) for name in _SPLITS + ["qrel"]
    ]
    
    def _info(self):

        if self.config.name in ["corpus", "topics"]:
             features=datasets.Features({
                "_id": datasets.Value("string"), 
                "title": datasets.Value("string"),
             })

        else:
            features = datasets.Features({
                    "query_id": datasets.Value("string"),
                    "doc_id": datasets.Value("string"),
                    "rel": datasets.Value("string")
                })
        return datasets.DatasetInfo(
            description=_DESCRIPTION,
            features=features,
            supervised_keys=None,
        )

    def _split_generators(self, dl_manager):
        """Returns SplitGenerators."""

        my_urls = _URLs[self.config.name]
        data_dir = dl_manager.download_and_extract(my_urls)

        return [
            datasets.SplitGenerator(
                name=self.config.name,
                # These kwargs will be passed to _generate_examples
                gen_kwargs={"filepath": data_dir},
            ),
        ]

    def _generate_examples(self, filepath):
        """Yields examples."""
        if self.config.name in ["corpus", "topics"]:
            with open(filepath, encoding="utf-8") as f:
                texts = f.readlines()
            for i, text in enumerate(texts):
                text = json.loads(text)
                if 'metadata' in text: del text['metadata']
                if "title" not in text: text["title"] = ""
                yield i, text
        else:
            with open(filepath, encoding="utf-8") as f:
                for line in f:
                    if "query-id" not in line:
                        fields = line.strip().split("\t")
                        query_id = fields[0]
                        doc_id = fields[1]
                        rel = int(fields[2])
                        yield query_id, doc_id, rel