miracl.py

"""MIRACL reranking dataset for multiple languages"""
import json
from typing import List, Dict

import datasets
from datasets import DatasetInfo

_DESCRIPTION = """
MIRACL (Multilingual Information Retrieval Across a Continuum of Languages) is a multilingual 
retrieval dataset that focuses on search across 18 different languages.
"""

_HOMEPAGE_URL = 'https://project-miracl.github.io/'
_LANGUAGES = ['es', 'de']
_VERSION = '1.0.0'
_URL = 'https://huggingface.co/datasets/jinaai/miracl/resolve/main/'
_URLS = {
    lang: _URL + f"{lang}-test.jsonl.gz" for lang in _LANGUAGES
}


class MIRACL(datasets.GeneratorBasedBuilder):
    def _info(self) -> DatasetInfo:
        features = datasets.Features(
            query=datasets.Value("string"),
            positive=datasets.Sequence(datasets.Value("string")),
            negative=datasets.Sequence(datasets.Value("string")),
        )
        return datasets.DatasetInfo(
            features=features,
            # Add any additional metadata about your dataset here
        )
    BUILDER_CONFIGS = [datasets.BuilderConfig(
            version=datasets.Version('1.0.0'),
            name=lang, description=f'MIRACL dataset in language {lang}.'
        ) for lang in _LANGUAGES
    ]

    def _split_generators(self, dl_manager: datasets.DownloadManager) -> List[datasets.SplitGenerator]:
        # Use the configuration name to select the language
        selected_lang = self.config.name
        if selected_lang not in _URLS:
            raise ValueError(f"Requested language '{selected_lang}' not found in the dataset.")

        file_path = dl_manager.download_and_extract(_URLS[selected_lang])
        return [datasets.SplitGenerator(name=datasets.Split.TEST, gen_kwargs={"filepath": file_path})]

    def _generate_examples(self, filepath: str) -> Dict[str, str]:
        # Load and yield examples from the jsonl file
        with open(filepath, encoding="utf-8") as f:
            for i, line in enumerate(f):
                # Parse the JSONL lines and yield examples
                json_line = json.loads(line)
                example = {
                    "query": json_line["query"],
                    "positive": json_line["positive"],
                    "negative": json_line["negative"],
                }
                yield str(i), example