search

RowitZou / CG-nAR

EMNLP-2021 paper: Thinking Clearly, Talking Fast: Concept-Guided Non-Autoregressive Generation for Open-Domain Dialogue Systems.
MIT License
18 stars 1 forks source link

How did you get file 'candi_keyword.txt'? #7

Closed unknowed-ER closed 2 years ago

unknowed-ER commented 2 years ago

You provide file 'src/preprocess/persona_process/convai2/candi_keyword.txt'. It's seem that this file is the topic keyword. How do you get file 'candi_keyword.txt'? How do I build this file if I use code to work with other datasets?

RowitZou commented 2 years ago

We follow the preprocessing from a previous work, which uses a rule-based method and PMI measuring to get keywords: https://github.com/squareRoot3/Target-Guided-Conversation