SETU contains 6,000 synthetic Hinglish and English e-commerce support
messages with ten structured labels. It is designed for supervised
fine-tuning and evaluation of small local language models.
Composition
6,000 total records.
4,800 train, 600 validation, 600 test.
50 scenario types.
1,348 template-family groups.
Seven intents: 2,600 complaint, 850 order inquiry, 800 refund request, 600
praise, 450… See the full description on the dataset page: https://huggingface.co/datasets/cyash1204/setu-hinglish-support-6000.
