Nebulixlabs/Saraswati-Hindi

Saraswati-Hindi

Saraswati-Hindi is an English-to-Hindi parallel text dataset containing automatically translated English sentences and their corresponding Hindi translations.

The dataset was created using the MyMemory Translation API to translate English text into Hindi (en → hi). It is intended for research, experimentation, and development of English-to-Hindi natural language processing (NLP) and machine translation systems.

Dataset Description

Overview

• —Dataset Name: Saraswati-Hindi
• —Language Pair: English → Hindi
• —Source Language: English (en)
• —Target Language: Hindi (hi)
• —Translation Method: Automatic machine translation
• —Translation API: MyMemory Translation API
• —Format: JSON
• —Task: Machine Translation / Parallel Text

Each record contains an English sentence and its corresponding Hindi translation.

Data Fields

The dataset contains the following fields:

Example

{
  "english_text": "Hello, how are you?",
  "translated_language": "hi",
  "translated_text": "नमस्ते, आप कैसे हैं?"
}

Dataset Creation

The dataset was generated using a Python script that reads English sentences from a text file and sends each sentence to the MyMemory Translation API.

The general pipeline is:

English Text
     ↓
Read sentence by sentence
     ↓
MyMemory Translation API
     ↓
English → Hindi translation
     ↓
Store translation
     ↓
JSON dataset

The translation process uses the language pair:

en|hi

A one-second delay was used between API requests to reduce the request rate.

Data Collection and Processing

The source English text was provided as a line-separated text file. Empty lines were skipped before translation.

For each non-empty line:

1. 1.The English sentence was read.
2. 2.The sentence was sent to the MyMemory Translation API.
3. 3.The API-generated Hindi translation was retrieved.
4. 4.The English sentence and Hindi translation were stored as a JSON record.
5. 5.The intermediate dataset was saved continuously to avoid losing previously processed translations.

Intended Use

Saraswati-Hindi may be useful for:

• —English-to-Hindi machine translation experiments
• —Hindi NLP research
• —Parallel corpus experimentation
• —Translation model prototyping
• —Multilingual language-model experiments
• —Sentence-level translation evaluation
• —Educational and research demonstrations

Limitations

This dataset is automatically translated and may contain:

• —Incorrect or awkward translations
• —Grammatical errors
• —Context-dependent translation issues
• —Literal translations
• —Ambiguous translations
• —Inconsistent terminology
• —Named-entity or formatting errors

The dataset has not necessarily been manually reviewed or validated by professional Hindi translators.

Therefore, users should perform additional cleaning and human evaluation before using the dataset for production systems or high-stakes applications.

Quality

Because the translations were generated automatically, translation quality can vary depending on the complexity and context of the English input.

Simple sentences are generally more suitable for this type of dataset, while long, technical, idiomatic, or context-dependent sentences may require additional verification.

Data Splits

The dataset currently contains the data generated by the provided translation pipeline.

No official train/validation/test split is assumed unless explicitly provided in the repository.

Users can create their own splits depending on their experimental requirements.

Licensing and Attribution

The English source texts and the resulting translations may be subject to the rights and licenses associated with the original source material and the translation service used to generate the data.

Before redistributing or using this dataset commercially, users should verify that they have the necessary rights to the original English source data and review the applicable MyMemory Translation API terms.

If you use this dataset in your research or project, please consider citing the dataset as:

@misc{saraswati-hindi,
  title        = {Saraswati-Hindi},
  author       = {Dataset Author},
  year         = {2026},
  publisher    = {Hugging Face},
  howpublished = {\url{https://huggingface.co/datasets/Nebulixlabs/Saraswati-Hindi}}
}

Acknowledgements

This dataset was generated with the help of the MyMemory Translation API.

We thank the developers and contributors of the translation service for providing the API used in the dataset generation pipeline.

Citation

If you use Saraswati-Hindi in your work, please cite the dataset using the citation information provided above.

Contact

For questions, suggestions, corrections, or dataset improvements, please use the Discussions or Issues section of the Hugging Face repository.

Dataset: Saraswati-Hindi Task: English → Hindi Machine Translation Languages: English (en) / Hindi (hi) Created with: MyMemory Translation API