Hugging Face's logo

Datasets:
mahtaa
/
trqa

Modalities:
Text
Formats:
json
Size:
100K - 1M
Libraries:
Datasets
pandas
Polars
License:
Dataset card Data Studio Files
xet
 Community
2
trqa / README.md
mahtaa's picture
mahtaa
HeydarS's picture
HeydarS
Add license (#2)
71e8c91
preview code
|
Raw History Blame Contribute Delete
2.39 kB
metadata 
configs:
  - config_name: queries
    data_files:
      - split: wiki1_test
        path: data/wiki1/test_queries_wiki1.jsonl
      - split: wiki1_validation
        path: data/wiki1/validation_queries_wiki1.jsonl
      - split: wiki2_test
        path: data/wiki2/test_queries_wiki2.jsonl
      - split: wiki2_validation
        path: data/wiki2/validation_queries_wiki2.jsonl
      - split: ecommerce_test
        path: data/ecommerce/test_queries_ecommerce.jsonl
      - split: ecommerce_validation
        path: data/ecommerce/validation_queries_ecommerce.jsonl
    default: true
  - config_name: qrels
    data_files:
      - split: wiki1_test
        path: data/wiki1/test_qrels_wiki1.jsonl
      - split: wiki1_validation
        path: data/wiki1/validation_qrels_wiki1.jsonl
      - split: wiki2_test
        path: data/wiki2/test_qrels_wiki2.jsonl
      - split: wiki2_validation
        path: data/wiki2/validation_qrels_wiki2.jsonl
      - split: ecommerce_test
        path: data/ecommerce/test_qrels_ecommerce.jsonl
      - split: ecommerce_validation
        path: data/ecommerce/validation_qrels_ecommerce.jsonl
  - config_name: corpus
    data_files:
      - split: wiki
        path: data/shared/corpus_wikipedia_en-20251001_rewritten.jsonl
      - split: wiki_partial
        path: data/shared/corpus_wikipedia_en-20251001_rewritten_partial.jsonl
      - split: ecommerce
        path: data/ecommerce/corpus_ecommerce_rewritten.jsonl
license: cc-by-sa-4.0

TRQA Dataset

TRQA dataset with three domains: wiki1, wiki2, and ecommerce.

Usage 

from datasets import load_dataset

# Load queries
queries = load_dataset("mahtaa/trqa", "queries", split="wiki1_test")

# Load qrels
qrels = load_dataset("mahtaa/trqa", "qrels", split="wiki1_test")

# Load corpus (wiki1 and wiki2 share the same Wikipedia corpus)
corpus = load_dataset("mahtaa/trqa", "corpus", split="wiki")

Configs and Splits

Config Splits Description
queries wiki1_test, wiki1_validation, wiki2_test, wiki2_validation, ecommerce_test, ecommerce_validation Questions with answers
qrels wiki1_test, wiki1_validation, wiki2_test, wiki2_validation, ecommerce_test, ecommerce_validation Relevance judgments (TREC format)
corpus wiki, wiki_partial,ecommerce Document collections

Note: wiki1 and wiki2 share the same Wikipedia corpus (wiki split for full corpus, or wiki_partial split for subsampled partial corpus).