Datasets:

Hyukkyu
/

beir-hotpotqa

license: apache-2.0
task_categories:
  - information-retrieval
  - text-retrieval
tags:
  - beir
  - hotpotqa
  - information-retrieval
  - retrieval
  - search
dataset_info:
  - config_name: corpus
    features:
      - name: title
        dtype: string
      - name: _id
        dtype: string
      - name: metadata
        struct:
          - name: url
            dtype: string
      - name: text
        dtype: string
    splits:
      - name: train
        num_bytes: 1870693670
        num_examples: 5233329
    download_size: 1010978786
    dataset_size: 1870693670
  - config_name: queries
    features:
      - name: text
        dtype: string
      - name: _id
        dtype: string
      - name: metadata
        dtype: string
    splits:
      - name: train
        num_bytes: 24804715
        num_examples: 97852
    download_size: 13448390
    dataset_size: 24804715
configs:
  - config_name: corpus
    data_files:
      - split: train
        path: corpus/train-*
  - config_name: queries
    data_files:
      - split: train
        path: queries/train-*

BEIR HOTPOTQA Dataset (Migrated)

This is a migrated version of BeIR/hotpotqa that is compatible with datasets library 4.0.0+.

Dataset Description

This dataset contains the hotpotqa dataset from the BEIR benchmark, converted from the old script-based format to Parquet format.

Dataset Structure

Queries

Split 'queries': 97,852 examples
- Features: ['_id', 'text', 'metadata']
Total examples: 97,852

Corpus

Split 'corpus': 5,233,329 examples
- Features: ['_id', 'title', 'text', 'metadata']
Total examples: 5,233,329

Usage

from datasets import load_dataset

# Load queries (split: queries)
queries = load_dataset("Hyukkyu/beir-hotpotqa", "queries", split="queries")

# Load corpus (split: corpus)
corpus = load_dataset("Hyukkyu/beir-hotpotqa", "corpus", split="corpus")

Available Splits

Queries

queries: 97,852 examples

Corpus

corpus: 5,233,329 examples

Original Dataset

This dataset is migrated from: BeIR/hotpotqa

Citation

If you use this dataset, please cite the original BEIR paper:

@article{thakur2021beir,
  title={BEIR: A Heterogeneous Benchmark for Zero-shot Evaluation of Information Retrieval Models},
  author={Thakur, Nandan and Reimers, Nils and Ruckle, Andreas and Srivastava, Abhishek and Gurevych, Iryna},
  journal={arXiv preprint arXiv:2104.08663},
  year={2021}
}