MUTANT

About 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

• Curated by: Lingo Research Group at IIT Gandhinagar
• Language(s) : Bilingual (Hindi [hi], English [en])
• Licensed by: cc-by-4.0

@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"
}