Hugging Face's logo

Datasets:
LingoIITGN
/
MUTANT

Tasks:
Summarization
Modalities:
Text
Formats:
text
Languages:
Hindi
English
Size:
100K - 1M
Tags:
code-mixing
text-summarization
Libraries:
Datasets
License:
Dataset card Data Studio Files Community
1
MUTANT / README.md
RajveeSheth's picture
RajveeSheth
Update README.md
a60dc62 verified
preview code
|
raw
history blame
2.57 kB
metadata
license: cc-by-4.0
language:
  - hi
  - en
tags:
  - code-mixing
  - text-summarization
size_categories:
  - 10K<n<100K
task_categories:
  - summarization
configs:
  - config_name: Jagran
    data_files:
      - split: train_part1
        path: Jagran/Part1/*.txt
      - split: train_part2
        path: Jagran/Part2/*.txt
      - split: train_part3
        path: Jagran/Part3/*.txt
      - split: train_part4
        path: Jagran/Part4/*.txt
      - split: train_part5
        path: Jagran/Part5/*.txt
  - config_name: Aap
    data_files:
      - split: train
        path: aap/*.txt
  - config_name: INC
    data_files:
      - split: train
        path: inc/*.txt
  - config_name: PIB
    data_files:
      - split: train
        path: pib/*.txt
  - config_name: Mankibaat
    data_files:
      - split: train
        path: mankibaat/*.txt
  - config_name: PM-speech
    data_files:
      - split: train
        path: pm-speech/*.txt

Dataset Overview

MUTANT (A Multi-sentential Code-mixed Hinglish Dataset): MUTANT is a high-quality Hindi-English code-mixed dataset designed for tasks related to multi-sentential text processing, particularly focusing on summarization and evaluation.

DATA Sources

MUTANT dataset comprises code-mixed long-length texts extracted from two main sources:

1. Political Speeches & Press Releases: Collected from government portals and political party websites.

2. Hindi News Articles: Extracted from leading Hindi news websites, ensuring high-quality and formal Hinglish text.

The final dataset contains multiple documents, each of which contains at least one code-mixed MCT. The dataset includes a total of 67007 documents with 84937 MCTs. A significant portion of the documents (44913) belong to the Dainik Jagran dataset.

Dataset Attribution & Licensing

Citation

If you use this dataset, please cite the following work:

@inproceedings{gupta-etal-2023-mutant,
    title = "{MUTANT}: A Multi-sentential Code-mixed {H}inglish Dataset",
    author = "Gupta, Rahul  and Srivastava, Vivek  and Singh, Mayank",
    editor = "Vlachos, Andreas  and
      Augenstein, Isabelle",
    booktitle = "Findings of the Association for Computational Linguistics: EACL 2023",
    month = may,
    year = "2023",
    address = "Dubrovnik, Croatia",
    publisher = "Association for Computational Linguistics",
    url = "https://aclanthology.org/2023.findings-eacl.56/",
    doi = "10.18653/v1/2023.findings-eacl.56",
    pages = "744--753"
}