Datasets:

Process-Venue
/

verb-translation-dataset

license: mit
task_categories:
  - translation
language:
  - en
  - hi
  - mr
  - sa
tags:
  - verb-translation
  - translation-pairs
  - multilingual-translation
  - english-hindi
  - english-marathi
  - english-sanskrit
size_categories:
  - 10K<n<100K

Process-Venue/verb-translation-dataset

Dataset Description

This dataset offers a comprehensive collection of verb translations from English into Hindi, Marathi, and Sanskrit, three major Indian languages. Each entry provides the English verb alongside its corresponding translations, enabling multilingual analysis and cross-lingual studies. The dataset is meticulously curated to support a wide range of natural language processing (NLP) and language technology applications, including but not limited to translation, language learning, and linguistic research. Its structured format and multilingual coverage make it a valuable resource for developing and benchmarking models in low-resource and morphologically rich languages.

Key features:

Parallel verb translations across four languages (English, Hindi, Marathi, Sanskrit)
Clean, structured format suitable for direct use in machine learning pipelines
Designed for both academic research and practical language technology development
Supports exploration of linguistic phenomena across Indo-Aryan languages

Dataset Summary

Total records: 15,953
Languages: English (source), Hindi, Marathi, Sanskrit (targets)
Columns: id, English Word, Hindi Translation, Marathi Translation, Sanskrit Translation, translation_pair
Format: Parquet

Dataset Structure

{
    'id': int,
    'English Word': str,
    'Hindi Translation': str,
    'Marathi Translation': str,
    'Sanskrit Translation': str,
    'translation_pair': str
}

Usage

from datasets import load_dataset

dataset = load_dataset("Process-Venue/verb-translation-dataset")

# Access the data
for example in dataset['train']:
    print(f"English: {example['English Word']}")
    print(f"Hindi: {example['Hindi Translation']}")
    print(f"Marathi: {example['Marathi Translation']}")
    print(f"Sanskrit: {example['Sanskrit Translation']}")

Data Fields

id: Unique identifier for each record
English Word: Source English verb/word
Hindi Translation: Hindi translation of the English word
Marathi Translation: Marathi translation of the English word
Sanskrit Translation: Sanskrit translation of the English word
translation_pair: Original text line from the source file

Dataset Statistics

Total records: 15,953
English words: 15,953
Hindi translations: 15,953
Marathi translations: 15,953
Sanskrit translations: 15,953

Languages

English (en): Source language
Hindi (hi): Target language (Devanagari script)
Marathi (mr): Target language (Devanagari script)
Sanskrit (sa): Target language (Devanagari script)

Applications

This dataset can be used for:

Multilingual Machine Translation: Train and evaluate models for translating verbs and short phrases between English, Hindi, Marathi, and Sanskrit.
Language Learning Tools: Develop educational apps and resources for learners of Indian languages, leveraging parallel verb lists for vocabulary acquisition and practice.
Cross-lingual Word Embeddings: Build and benchmark word embedding models that capture semantic relationships across multiple languages.
Low-resource NLP Model Training: Support the creation of NLP models for Hindi, Marathi, and Sanskrit, which are often underrepresented in large-scale datasets.
Data Augmentation for AI: Enrich other datasets or tasks (e.g., question answering, paraphrasing) with high-quality, aligned verb translations.
Script Conversion and Transliteration: Develop and test tools for converting between Devanagari and Latin scripts, or for transliterating between languages.

Citation

If you use this dataset, please cite:

@dataset{verb_translation_dataset,
  title={Multilingual Verb Translation Dataset: English to Hindi, Marathi, and Sanskrit},
  author={processVenue},
  year={2025},
  url={https://huggingface.co/datasets/Process-Venue/verb-translation-dataset}
}

License

This dataset is released under the MIT License.