README.md

metadata

license: cc-by-sa-3.0
configs:
  - config_name: ara_Arab
    data_files:
      - split: validation
        path: data/ara_Arab/validation*
      - split: test
        path: data/ara_Arab/test*
  - config_name: deu_Latn
    data_files:
      - split: validation
        path: data/deu_Latn/validation*
      - split: test
        path: data/deu_Latn/test*
  - config_name: eng_Latn
    data_files:
      - split: validation
        path: data/eng_Latn/validation*
      - split: test
        path: data/eng_Latn/test*
  - config_name: hin_Deva
    data_files:
      - split: validation
        path: data/hin_Deva/validation*
      - split: test
        path: data/hin_Deva/test*
  - config_name: hin_Latn
    data_files:
      - split: validation
        path: data/hin_Latn/validation*
      - split: test
        path: data/hin_Latn/test*
  - config_name: spa_Latn
    data_files:
      - split: validation
        path: data/spa_Latn/validation*
      - split: test
        path: data/spa_Latn/test*
  - config_name: vie_Latn
    data_files:
      - split: validation
        path: data/vie_Latn/validation*
      - split: test
        path: data/vie_Latn/test*
  - config_name: zho_Hans
    data_files:
      - split: validation
        path: data/zho_Hans/validation*
      - split: test
        path: data/zho_Hans/test*
task_categories:
  - question-answering
language:
  - en
  - hi
  - ar
  - de
  - es
  - vi
  - zh

Source Dataset

• Link: facebook/mlqa
• Revision: 397ed406c1a7902140303e7faf60fff35b58d285

MLQA
MLQA (MultiLingual Question Answering) is a benchmark dataset for evaluating cross-lingual question answering performance. MLQA consists of over 5K extractive QA instances (12K in English) in SQuAD format in seven languages - English, Arabic, German, Spanish, Hindi, Vietnamese and Simplified Chinese. MLQA is highly parallel, with QA instances parallel between 4 different languages on average.

MLQA Plus
MLQA Plus additionally has hin_Latn data generated using indictrans library.