README.md

metadata

extra_gated_prompt: You agree to use this dataset solely for non-commercial and research purposes.
extra_gated_fields:
  Country: country
  Specific date: date_picker
  I want to use this dataset for:
    type: select
    options:
      - Research
      - Education
      - label: Other
        value: other
  I agree to use this dataset for non-commercial use ONLY: checkbox
configs:
  - config_name: wiki
    data_files:
      - split: train
        path: wiki_odia_253Ks_4p11Mw.json
  - config_name: oscar
    data_files:
      - split: train
        path: oscar_odia_1p16Ms_25Mw.json
  - config_name: paraphrasing
    data_files:
      - split: train
        path: paraphrasing_odia_105Ks_2p3Mw.json
  - config_name: indicQA
    data_files:
      - split: train
        path: indicQA_odia_12Ks_184Kw.json
  - config_name: sentiment_analysis
    data_files:
      - split: train
        path: sentiment_analysis_odia_1Ks_34Kw.json
  - config_name: odiaencorp
    data_files:
      - split: train
        path: odiaencorp_85Ks_1p1Mw.json
  - config_name: xp3
    data_files:
      - split: train
        path: xp3_261Ks_4p9Mw.json
  - config_name: samanantar
    data_files:
      - split: train
        path: samanantar_odia_990Ks_10Mw.json
  - config_name: cultureax
    data_files:
      - split: train
        path: cultureax_odia_2p9Ks_49Mw.json
  - config_name: pmo
    data_files:
      - split: train
        path: pmo_data.json
  - config_name: varta
    data_files:
      - split: validation
        path: val_or_shard_01.json
      - split: train
        path: train_or_shard_01.json
size_categories:
  - 1M<n<10M

The present dataset is compiled by using the following datasets:

CultureaX

• Licesnse - ODC-By, CC0, Paper
• Source - https://huggingface.co/datasets/uonlp/CulturaX/viewer/or?
• 49M tokens, 2.9M sentences
• Collection of different versions of Ocsar (Commom Crawl data) and mC4 dataset (Common Crawl's web crawl corpus). mC4 forms 66% of CulturaX dataset.

IndicQA

• License - cc-by-4.0
• Source - https://huggingface.co/datasets/ai4bharat/IndicQA/viewer/indicqa.or
• 0.23M tokens, 15K sentences
• In-context question-answering in Indic languages.

Odiaencorp

• License - CC BY-NC-SA 4.0
• Source - https://lindat.mff.cuni.cz/repository/xmlui/handle/11234/1-3211
• 1.1M tokens, 85K sentences
• Odia wiki, PM India, Odia Digital library, Odisha Govt. websites, Bilingual literature books.

Oscar

• License - cc0-1.0, Paper
• Source - https://huggingface.co/datasets/oscar-corpus/OSCAR-2201
• 25.7M tokens, 1.2M sentences
• Common crawl data.

Paraphrasing

• License - cc-by-nc-4.0, Paper
• Source - https://huggingface.co/datasets/ai4bharat/IndicParaphrase/viewer/or
• 2.3M tokens, 0.105M sentences
• Rephrasing a given sentence/paragraph.

PMO

• License - cc-by-nc-4.0
• Source - PMO website
• 2.2M tokens, 0.131M sentences
• Speech, Statements, Activities of PM of India.

Samanantar

• License - cc-by-nc-4.0, Paper
• Source - https://huggingface.co/datasets/ai4bharat/samanantar/viewer/or
• 10.25M Tokens, 0.909M Sentences
• Collection of short sentences.

Sentiment Analysis

• License - Unspecified
• Source - https://huggingface.co/datasets/ai4bharat/IndicSentiment/viewer/translation-or
• 34K Tokens, 1000 Sentences
• Generic sentences and their sentiment analysis.

Varta

• License - cc
• Source - https://huggingface.co/datasets/rahular/varta
• 193.5M tokens, 14.19M sentences.
• News articles.

Wiki

• License - cc-by-sa-3.0, gfdl
• Source - https://huggingface.co/datasets/wikimedia/wikipedia/viewer/20231101.or
• 4.1M tokens, 0.253M sentences
• Wikipedia Articles.

XP3

• License - apache-2.0
• Source - https://huggingface.co/datasets/bigscience/xP3all
• 4.9M tokens, 0.26M sentences
• Generic statements machine-translated into Odia.