mahendraphd/Indic_Hindi-English_Parallel_Speech

---

📘 TSV File Format

Each split (train.tsv, validation.tsv, test.tsv) follows this 4-column format:

hi_audio	en_audio	hi_text	en_text
hi_audio/0001_hi.wav	en_audio/0001_en.wav	नमस्ते आप कैसे हैं?	Hello, how are you?

Column Description

• hi_audio → Hindi audio file path
• en_audio → English audio file path
• hi_text → Hindi text transcription
• en_text → English text transcription

---

🧠 Use Cases

This dataset is suitable for:

🔊 Speech-to-Speech Translation (S2ST)

• Translatotron
• S2UT
• SeamlessM4T

🎧 Speech-to-Text Translation (ST)

• Whisper
• S2T Transformer
• mSLAM / HuBERT / Wav2Vec2

🔤 Multimodal Learning

• Q-Former training
• Speech–text contrastive learning
• Multilingual embedding learning

---

📊 Dataset Statistics

Split	#Samples	Notes
Train	108364	Majority of the data
Validation	6059	For tuning
Test	6052	Final evaluation

(Update the counts once finalized.)

---

🔧 Load with Hugging Face Datasets

from datasets import load_dataset

dataset = load_dataset("mahendraphd/Indic_Hindi-English_Parallel_Speech")

print(dataset["train"][0])

📄 Publication

This dataset accompanies the research paper:

“Benchmarking Hindi-to-English direct speech-to-speech translation with synthetic data”
Gupta, Mahendra and Dutta, Maitreyee and Maurya, Chandresh Kumar
(Language Resources and Evaluation – Published)

If you use this dataset, please cite both the dataset and the paper.

BibTeX

@article{gupta2025benchmarking,
  title={Benchmarking Hindi-to-English direct speech-to-speech translation with synthetic data},
  author={Gupta, Mahendra and Dutta, Maitreyee and Maurya, Chandresh Kumar},
  journal={Language Resources and Evaluation},
  pages={1--39},
  year={2025},
  publisher={Springer}
}