Dataset Summary
This dataset is a large-scale English–Khasi parallel corpus consisting of approximately 340,000 sentence pairs. It has been constructed to support research and development in Neural Machine Translation (NMT), low-resource language modeling, and cross-lingual representation learning involving Khasi.
The dataset significantly expands existing English–Khasi resources by incorporating synthetic parallel data alongside curated sentence structures, enabling improved coverage of linguistic patterns and vocabulary.
Supported Tasks
- Machine Translation (English ↔ Khasi)
- Cross-lingual Transfer Learning
- Language Modeling for Low-Resource Languages
- Data Augmentation for NMT systems
- Evaluation Benchmarks for Khasi translation systems
Languages
- English (en)
- Khasi (kha)
Limitations
Some translations may exhibit:
- Reduced fluency
- Literal phrasing
- Minor grammatical inconsistencies
- Manual verification was only done for 20K pairs as I am working on this alone
- Potential bias due to synthetic data generation as back-translation was done using my previous translation models
- Downloads last month
- 34