This dataset aggregates approximately 50,000 sentence pairs and over 318,000 words of English-Hindi parallel text across news, literature, and technical domains. It incorporates manually verified alignments from the Intercorp literary corpus and specialized sub-corpora for agriculture and Wikipedia named entities.
Use Cases
- Train English-to-Hindi translation models using the sentence-aligned pairs from the Tides and Intercorp sub-corpora
- Build Hindi named-entity recognition (NER) tools using the Wikipedia-crawled entity list
- Perform domain adaptation for machine translation using the specialized agriculture parallel corpus
Strengths
- 50,000 sentence pairs sourced from the DARPA-TIDES news article collection
- 118,000 Hindi words from Intercorp featuring manually checked sentence alignments
- 200,000 words of parallel text from the EMILLE corpus including South Asian languages
- Domain-specific parallel text covering agriculture and Wikipedia-sourced named entities