This is a combined dataset of Part 1 and Part 3 of IMDA NSC.

- Train set: 100% P1, 90% P2
- Train-dev set: 5% P2
- Test Set: 5% P2

`input_ids` are generated using [JET2001/singlish-tokenizer](https://huggingface.co/JET2001/singlish-tokenizer)