We prompted ChatGPT (Feb 13, 2023 version), InstructGPT (davinci-002 and davinci-003), BLOOMZ and Flan-T5-XXL with six different prompt templates in a zero-shot fashion to generate code-mixed sentences for five different topics and six South East Asian languages (Malay, Indonesian, Chinese, Tagalog, Vietnamese, and Singlish).
The data
folder contains tsv (tab-separated values) files for our annotations.