Datasets:

Modalities:
Text
Formats:
parquet
Languages:
Persian
ArXiv:
Libraries:
Datasets
Dask
License:
FarsInstruct / README.md
hojjat-m's picture
Upload dataset
c066903 verified
metadata
language:
  - fa
license: apache-2.0
task_categories:
  - text-classification
  - question-answering
  - translation
  - text-generation
pretty_name: FarsInstruct
dataset_info:
  - config_name: default
    features:
      - name: inputs
        dtype: string
      - name: outputs
        dtype: string
      - name: dataset
        dtype: string
      - name: template
        dtype: string
    splits:
      - name: train
        num_bytes: 6564618613
        num_examples: 9374312
      - name: validation
        num_bytes: 330008882
        num_examples: 316117
      - name: test
        num_bytes: 902009657
        num_examples: 1308596
    download_size: 3595052926
    dataset_size: 7796637152
  - config_name: digi_sentiment
    features:
      - name: inputs
        dtype: string
      - name: outputs
        dtype: string
      - name: dataset
        dtype: string
      - name: template
        dtype: string
    splits:
      - name: train
        num_bytes: 17608449
        num_examples: 22820
      - name: validation
        num_bytes: 3882215
        num_examples: 4890
      - name: test
        num_bytes: 3868362
        num_examples: 4900
    download_size: 8359417
    dataset_size: 25359026
  - config_name: digimag
    features:
      - name: inputs
        dtype: string
      - name: outputs
        dtype: string
      - name: dataset
        dtype: string
      - name: template
        dtype: string
    splits:
      - name: train
        num_bytes: 238365098
        num_examples: 68960
      - name: validation
        num_bytes: 26780346
        num_examples: 7670
      - name: test
        num_bytes: 28913146
        num_examples: 8520
    download_size: 133522898
    dataset_size: 294058590
  - config_name: exappc
    features:
      - name: inputs
        dtype: string
      - name: outputs
        dtype: string
      - name: dataset
        dtype: string
      - name: template
        dtype: string
    splits:
      - name: train
        num_bytes: 153837081
        num_examples: 270269
      - name: validation
        num_bytes: 32613089
        num_examples: 57592
      - name: test
        num_bytes: 33348597
        num_examples: 58033
    download_size: 81725354
    dataset_size: 219798767
  - config_name: farstail
    features:
      - name: inputs
        dtype: string
      - name: outputs
        dtype: string
      - name: dataset
        dtype: string
      - name: template
        dtype: string
    splits:
      - name: train
        num_bytes: 55160519
        num_examples: 72660
      - name: validation
        num_bytes: 11531714
        num_examples: 15370
      - name: test
        num_bytes: 11743134
        num_examples: 15640
    download_size: 29639181
    dataset_size: 78435367
  - config_name: p3_qa_translated
    features:
      - name: inputs
        dtype: string
      - name: outputs
        dtype: string
      - name: dataset
        dtype: string
      - name: template
        dtype: string
    splits:
      - name: train
        num_bytes: 369861273
        num_examples: 646505
      - name: validation
        num_bytes: 39799885
        num_examples: 76565
    download_size: 114937546
    dataset_size: 409661158
  - config_name: pars_absa
    features:
      - name: inputs
        dtype: string
      - name: outputs
        dtype: string
      - name: dataset
        dtype: string
      - name: template
        dtype: string
    splits:
      - name: train
        num_bytes: 42114471
        num_examples: 45006
      - name: validation
        num_bytes: 7177753
        num_examples: 7506
      - name: test
        num_bytes: 6919680
        num_examples: 7500
    download_size: 20744680
    dataset_size: 56211904
  - config_name: parsinlu_comp
    features:
      - name: inputs
        dtype: string
      - name: outputs
        dtype: string
      - name: dataset
        dtype: string
      - name: template
        dtype: string
    splits:
      - name: train
        num_bytes: 7262540
        num_examples: 6000
      - name: validation
        num_bytes: 1587072
        num_examples: 1250
      - name: test
        num_bytes: 6613589
        num_examples: 5700
    download_size: 6217942
    dataset_size: 15463201
  - config_name: parsinlu_en_fa
    features:
      - name: inputs
        dtype: string
      - name: outputs
        dtype: string
      - name: dataset
        dtype: string
      - name: template
        dtype: string
    splits:
      - name: train
        num_bytes: 954515437
        num_examples: 3000000
      - name: validation
        num_bytes: 4644558
        num_examples: 12822
      - name: test
        num_bytes: 137665422
        num_examples: 290154
    download_size: 371019560
    dataset_size: 1096825417
  - config_name: parsinlu_entailment
    features:
      - name: inputs
        dtype: string
      - name: outputs
        dtype: string
      - name: dataset
        dtype: string
      - name: template
        dtype: string
    splits:
      - name: train
        num_bytes: 5633585
        num_examples: 8296
      - name: validation
        num_bytes: 2057795
        num_examples: 2970
      - name: test
        num_bytes: 12628240
        num_examples: 18407
    download_size: 6613453
    dataset_size: 20319620
  - config_name: parsinlu_fa_en
    features:
      - name: inputs
        dtype: string
      - name: outputs
        dtype: string
      - name: dataset
        dtype: string
      - name: template
        dtype: string
    splits:
      - name: train
        num_bytes: 951650636
        num_examples: 2999995
      - name: validation
        num_bytes: 9269774
        num_examples: 25644
      - name: test
        num_bytes: 267276146
        num_examples: 572928
    download_size: 429940819
    dataset_size: 1228196556
  - config_name: parsinlu_multiple_choice
    features:
      - name: inputs
        dtype: string
      - name: outputs
        dtype: string
      - name: dataset
        dtype: string
      - name: template
        dtype: string
    splits:
      - name: train
        num_bytes: 5569218
        num_examples: 12710
      - name: validation
        num_bytes: 229444
        num_examples: 417
      - name: test
        num_bytes: 4109446
        num_examples: 9450
    download_size: 2907686
    dataset_size: 9908108
  - config_name: parsinlu_qpp
    features:
      - name: inputs
        dtype: string
      - name: outputs
        dtype: string
      - name: dataset
        dtype: string
      - name: template
        dtype: string
    splits:
      - name: train
        num_bytes: 8575019
        num_examples: 17164
      - name: validation
        num_bytes: 4246069
        num_examples: 8418
      - name: test
        num_bytes: 8529281
        num_examples: 18078
    download_size: 5262177
    dataset_size: 21350369
  - config_name: parsinlu_sentiment
    features:
      - name: inputs
        dtype: string
      - name: outputs
        dtype: string
      - name: dataset
        dtype: string
      - name: template
        dtype: string
    splits:
      - name: train
        num_bytes: 73453659
        num_examples: 136170
    download_size: 19197308
    dataset_size: 73453659
  - config_name: persian_ner
    features:
      - name: inputs
        dtype: string
      - name: outputs
        dtype: string
      - name: dataset
        dtype: string
      - name: template
        dtype: string
    splits:
      - name: train
        num_bytes: 79721941
        num_examples: 143388
      - name: test
        num_bytes: 39461389
        num_examples: 71680
    download_size: 42457467
    dataset_size: 119183330
  - config_name: persian_news
    features:
      - name: inputs
        dtype: string
      - name: outputs
        dtype: string
      - name: dataset
        dtype: string
      - name: template
        dtype: string
    splits:
      - name: train
        num_bytes: 215514328
        num_examples: 79884
      - name: validation
        num_bytes: 23986039
        num_examples: 8880
      - name: test
        num_bytes: 26811003
        num_examples: 9864
    download_size: 115002364
    dataset_size: 266311370
  - config_name: persian_qa
    features:
      - name: inputs
        dtype: string
      - name: outputs
        dtype: string
      - name: dataset
        dtype: string
      - name: template
        dtype: string
    splits:
      - name: train
        num_bytes: 173326369
        num_examples: 81974
      - name: validation
        num_bytes: 17626045
        num_examples: 8463
    download_size: 84699623
    dataset_size: 190952414
  - config_name: peyma
    features:
      - name: inputs
        dtype: string
      - name: outputs
        dtype: string
      - name: dataset
        dtype: string
      - name: template
        dtype: string
    splits:
      - name: train
        num_bytes: 44356046
        num_examples: 80280
      - name: validation
        num_bytes: 4981884
        num_examples: 9250
      - name: test
        num_bytes: 5799753
        num_examples: 10260
    download_size: 18549315
    dataset_size: 55137683
  - config_name: pn_sum
    features:
      - name: inputs
        dtype: string
      - name: outputs
        dtype: string
      - name: dataset
        dtype: string
      - name: template
        dtype: string
    splits:
      - name: train
        num_bytes: 2037812524
        num_examples: 902242
      - name: validation
        num_bytes: 20817174
        num_examples: 11184
      - name: test
        num_bytes: 138618064
        num_examples: 61523
    download_size: 947261646
    dataset_size: 2197247762
  - config_name: snapp_sentiment
    features:
      - name: inputs
        dtype: string
      - name: outputs
        dtype: string
      - name: dataset
        dtype: string
      - name: template
        dtype: string
    splits:
      - name: train
        num_bytes: 264483001
        num_examples: 573210
      - name: validation
        num_bytes: 42179746
        num_examples: 91707
      - name: test
        num_bytes: 45729114
        num_examples: 99363
    download_size: 93188564
    dataset_size: 352391861
  - config_name: syntran
    features:
      - name: inputs
        dtype: string
      - name: outputs
        dtype: string
      - name: dataset
        dtype: string
      - name: template
        dtype: string
    splits:
      - name: train
        num_bytes: 240609866
        num_examples: 481060
    download_size: 66613727
    dataset_size: 240609866
  - config_name: wiki_sum
    features:
      - name: inputs
        dtype: string
      - name: outputs
        dtype: string
      - name: dataset
        dtype: string
      - name: template
        dtype: string
    splits:
      - name: train
        num_bytes: 1075245114
        num_examples: 365224
      - name: validation
        num_bytes: 120120218
        num_examples: 40584
      - name: test
        num_bytes: 133195903
        num_examples: 45096
    download_size: 629544114
    dataset_size: 1328561235
  - config_name: xl_wic
    features:
      - name: inputs
        dtype: string
      - name: outputs
        dtype: string
      - name: dataset
        dtype: string
      - name: template
        dtype: string
    splits:
      - name: train
        num_bytes: 2773592
        num_examples: 7000
      - name: validation
        num_bytes: 588034
        num_examples: 1500
      - name: test
        num_bytes: 594874
        num_examples: 1500
    download_size: 1404098
    dataset_size: 3956500
configs:
  - config_name: default
    data_files:
      - split: train
        path: data/train-*
      - split: validation
        path: data/validation-*
      - split: test
        path: data/test-*
  - config_name: digi_sentiment
    data_files:
      - split: train
        path: digi_sentiment/train-*
      - split: validation
        path: digi_sentiment/validation-*
      - split: test
        path: digi_sentiment/test-*
  - config_name: digimag
    data_files:
      - split: train
        path: digimag/train-*
      - split: validation
        path: digimag/validation-*
      - split: test
        path: digimag/test-*
  - config_name: exappc
    data_files:
      - split: train
        path: exappc/train-*
      - split: validation
        path: exappc/validation-*
      - split: test
        path: exappc/test-*
  - config_name: farstail
    data_files:
      - split: train
        path: farstail/train-*
      - split: validation
        path: farstail/validation-*
      - split: test
        path: farstail/test-*
  - config_name: p3_qa_translated
    data_files:
      - split: train
        path: p3_qa_translated/train-*
      - split: validation
        path: p3_qa_translated/validation-*
  - config_name: pars_absa
    data_files:
      - split: train
        path: pars_absa/train-*
      - split: validation
        path: pars_absa/validation-*
      - split: test
        path: pars_absa/test-*
  - config_name: parsinlu_comp
    data_files:
      - split: train
        path: parsinlu_comp/train-*
      - split: validation
        path: parsinlu_comp/validation-*
      - split: test
        path: parsinlu_comp/test-*
  - config_name: parsinlu_en_fa
    data_files:
      - split: train
        path: parsinlu_en_fa/train-*
      - split: validation
        path: parsinlu_en_fa/validation-*
      - split: test
        path: parsinlu_en_fa/test-*
  - config_name: parsinlu_entailment
    data_files:
      - split: train
        path: parsinlu_entailment/train-*
      - split: validation
        path: parsinlu_entailment/validation-*
      - split: test
        path: parsinlu_entailment/test-*
  - config_name: parsinlu_fa_en
    data_files:
      - split: train
        path: parsinlu_fa_en/train-*
      - split: validation
        path: parsinlu_fa_en/validation-*
      - split: test
        path: parsinlu_fa_en/test-*
  - config_name: parsinlu_multiple_choice
    data_files:
      - split: train
        path: parsinlu_multiple_choice/train-*
      - split: validation
        path: parsinlu_multiple_choice/validation-*
      - split: test
        path: parsinlu_multiple_choice/test-*
  - config_name: parsinlu_qpp
    data_files:
      - split: train
        path: parsinlu_qpp/train-*
      - split: validation
        path: parsinlu_qpp/validation-*
      - split: test
        path: parsinlu_qpp/test-*
  - config_name: parsinlu_sentiment
    data_files:
      - split: train
        path: parsinlu_sentiment/train-*
  - config_name: persian_ner
    data_files:
      - split: train
        path: persian_ner/train-*
      - split: test
        path: persian_ner/test-*
  - config_name: persian_news
    data_files:
      - split: train
        path: persian_news/train-*
      - split: validation
        path: persian_news/validation-*
      - split: test
        path: persian_news/test-*
  - config_name: persian_qa
    data_files:
      - split: train
        path: persian_qa/train-*
      - split: validation
        path: persian_qa/validation-*
  - config_name: peyma
    data_files:
      - split: train
        path: peyma/train-*
      - split: validation
        path: peyma/validation-*
      - split: test
        path: peyma/test-*
  - config_name: pn_sum
    data_files:
      - split: train
        path: pn_sum/train-*
      - split: validation
        path: pn_sum/validation-*
      - split: test
        path: pn_sum/test-*
  - config_name: snapp_sentiment
    data_files:
      - split: train
        path: snapp_sentiment/train-*
      - split: validation
        path: snapp_sentiment/validation-*
      - split: test
        path: snapp_sentiment/test-*
  - config_name: syntran
    data_files:
      - split: train
        path: syntran/train-*
  - config_name: wiki_sum
    data_files:
      - split: train
        path: wiki_sum/train-*
      - split: validation
        path: wiki_sum/validation-*
      - split: test
        path: wiki_sum/test-*
  - config_name: xl_wic
    data_files:
      - split: train
        path: xl_wic/train-*
      - split: validation
        path: xl_wic/validation-*
      - split: test
        path: xl_wic/test-*

Dataset Description

Dataset Summary

Instruction-tuned large language models, such as T0, have demonstrated remarkable capabilities in following instructions across various domains. However, their proficiency remains notably deficient in many low-resource languages. To address this challenge, we introduce FarsInstruct: a comprehensive instruction dataset designed to enhance the instruction-following ability of large language models specifically for the Persian language, a significant yet underrepresented language globally. FarsInstruct encompasses a wide range of task types and datasets, each containing a mix of straightforward to complex manual written instructions, as well as translations from Public Pool of Prompts, ensuring a rich linguistic and cultural representation. As of the current writing, FarsInstruct comprises more than 200 templates across 21 distinct datasets, and we intend to update it consistently, thus augmenting its applicability.

Supported tasks

Figure 1: Detailed depiction of 11 task types utilized in our dataset. Each box within the figure lists the specific datasets associated with the respective task type. Datasets designated for training are highlighted in blue, and those reserved for testing are marked in orange. Additionally, manual datasets, which have been specifically curated and prompted by our team, are enclosed with solid borders. In contrast, datasets that have been translated from English to Persian are enclosed with dashed borders

Citation information

@misc{mokhtarabadi2024farsinstructempoweringlargelanguage,
      title={FarsInstruct: Empowering Large Language Models for Persian Instruction Understanding}, 
      author={Hojjat Mokhtarabadi and Ziba Zamani and Abbas Maazallahi and Hossein Manshaei},
      year={2024},
      eprint={2407.11186},
      archivePrefix={arXiv},
      primaryClass={cs.CL},
      url={https://arxiv.org/abs/2407.11186}, 
}