The dataset viewer is not available because its heuristics could not detect any supported data files. You can try uploading some data files, or configuring the data files location manually.
📘 IL-ILGOV-2024: Indian Language Translation Benchmark
IL-ILGOV-2024 is a multilingual translation benchmark corpus for Indian languages developed under the HIndustani Machini ANuvaad TechnoloGY (HIMANGY) / MTIL project, with Hindi as source language and 12 target languages in the governance domain. (ResearchGate)
This dataset is aimed at evaluating and benchmarking machine translation (MT) systems across Indian languages in a domain that reflects real-world governance text.
📦 Repository Contents
The repository contains:
nway_parallel.asm
nway_parallel.ban
nway_parallel.eng
nway_parallel.guj
nway_parallel.hin
nway_parallel.kan
nway_parallel.mal
nway_parallel.mar
nway_parallel.odi
nway_parallel.pan
nway_parallel.tam
nway_parallel.tel
nway_parallel.urd
Each file represents N-way parallel sentence sets for Indian language translation.
⚠️ Note: The repository currently lacks a descriptive README and metadata; this document serves to fill that gap.
📖 About the Dataset
🧠 Purpose
The IL-ILGOV-2024 corpus was created to provide a controlled benchmarking resource to assess the performance of MT systems translating from Hindi to various Indian languages. (Springer)
🌐 Languages Covered
The corpus spans 12 languages:
| Code | Language |
|---|---|
| hin | Hindi |
| ass | Assamese |
| ben | Bangla |
| eng | English |
| guj | Gujarati |
| kan | Kannada |
| mal | Malayalam |
| mar | Marathi |
| odi | Odia |
| pan | Punjabi |
| tam | Tamil |
| tel | Telugu |
| urd | Urdu |
(Note: language codes reflect the file prefixes in this repository.)
📊 Data Structure
The benchmark corpus includes:
1304 n-way parallel sentences across all languages.
Three bilingual splits:
devdevtesttestEach split contains between 1000 and 3000 sentence pairs for Hindi ↔ target language testing. (ResearchGate)
🧾 Domain and Source
Domain: Governance (policy, administrative text).
Sentences were:
- Sourced online
- Automatically aligned
- Human-validated and corrected for alignment and translation quality to ensure high reliability for benchmarking. (ResearchGate)
🚀 Use Cases
This benchmark can be used for:
- Evaluating MT models across Indian language pairs.
- Comparative performance analysis (e.g., BLEU, COMET).
- Training and testing multilingual MT systems especially for low-resource Indian languages.
- Domain-specific MT research in governance.
📄 Published Paper
The dataset and methodology are described in the peer-reviewed article:
Title: IL-ILGOV-2024: a translation benchmark for Hindi-to-12 languages in the governance domain Authors: Vandan Mujadia, Rao B. Ashwath, Dipti Misra Sharma Journal: Language Resources and Evaluation (2025) Published: 06 September 2025 DOI: https://doi.org/10.1007/s10579-025-09859-8 (Springer)
This paper discusses:
- Purpose and motivation for the benchmark.
- Dataset creation pipeline including alignment and human validation.
- Benchmark statistics and structure.
- Utility for MT system evaluation.
📝 Citation
If you use this dataset or benchmark in your research, please cite the paper:
@article{Mujadia2025ILILGOV,
title={IL-ILGOV-2024: a translation benchmark for Hindi-to-12 languages in the governance domain},
author={Mujadia, Vandan and Ashwath, Rao B. and Misra Sharma, Dipti},
journal={Language Resources and Evaluation},
year={2025},
pages={3851--3872},
doi={10.1007/s10579-025-09859-8}
}
📬 Contact & Contributions
For questions or contributions, you may:
- Open an issue or pull request on the GitHub repository.
- Contact the dataset authors (see the paper for email addresses).
🧪 License
Please check the original repository or paper for licensing details. The current GitHub repo doesn’t specify a license yet.
If you want, I can also generate sample evaluation scripts (e.g., for BLEU/COMET scoring) or annotate the data format for easier integration with MT toolkits.
- Downloads last month
- 27