agnivamaiti/naganlp-conversational-corpus

NagaNLP Conversational Corpus

Dataset Summary

This is the official conversational dataset for the NagaNLP project. It contains 10,021 instruction-following pairs (User / Assistant) in Nagamese (Naga Pidgin), intended for instruction-tuning or evaluating conversational models in a low-resource creole language.

Supported Tasks

• —Text Generation / Instruction Following: given a user prompt in Nagamese, generate an appropriate assistant response in Nagamese.

Languages

Nagamese / Naga Pidgin (ISO 639-3: nag), a creole widely used as a lingua franca across Nagaland and neighboring Northeast Indian states.

Dataset Structure

Data Instance

{
  "prompt": "User: Itu kath laga 'tobacco pipe' bhal lagishe.",
  "completion": "Assistant: Hoi, purana manu khan e..."
}

Data Fields

• —prompt: string — the user's instruction, prefixed with User:.
• —completion: string — the assistant's response, prefixed with Assistant:.

Data Splits

Note: This is the full, unsplit corpus. It is mapped to the train split purely for HuggingFace Dataset Viewer compatibility — there is no dedicated validation/test split.

Dataset Creation

Source Data

Instruction/response pairs were constructed as part of the NagaNLP project's human-in-the-loop synthetic data pipeline, covering everyday conversational scenarios (greetings, requests, small talk, general Q&A) in Nagamese.

Annotations

Pairs were generated and verified through a human-in-the-loop process; see the accompanying NagaNLP paper for full methodology.

Considerations for Using the Data

Limitations

• —As with other NagaNLP resources, a portion of this corpus originates from a synthetic, human-in-the-loop generation pipeline rather than organically collected conversations — this should be disclosed when the dataset is used to train or evaluate models intended to reflect naturalistic Nagamese usage.
• —Coverage of conversational domains (topics, registers, dialectal variation within Nagamese) is not exhaustive.

Licensing

Licensed under CC BY-NC 4.0 (attribution required, non-commercial use only). For commercial licensing inquiries, please contact the author.

Citation

@misc{maiti2025naganlp,
      title={NagaNLP: Bootstrapping NLP for Low-Resource Nagamese Creole with Human-in-the-Loop Synthetic Data},
      author={Agniva Maiti and Manya Pandey and Murari Mandal},
      year={2025},
      eprint={2512.12537},
      archivePrefix={arXiv},
      primaryClass={cs.CL},
      url={https://arxiv.org/abs/2512.12537}
}

Contact

Agniva Maiti — KIIT University — HuggingFace: agnivamaiti