About the code of the evaluation part

thunlp / SememePSO-Attack

Code and data of the ACL 2020 paper "Word-level Textual Adversarial Attacking as Combinatorial Optimization"

MIT License

86 stars 14 forks source link

About the code of the evaluation part #5

Closed milumilum closed 4 years ago

milumilum commented 4 years ago

Hi, could you please pull up the code of the evaluation part? I am also doing relevant experiments, but the sentence complexity is quite different from your results, so I make this request and I want to learn from your code. thanks.

yangalan123 commented 4 years ago

Hi! Thanks for your interest. I am one of the main maintainers of this project. I am a bit confused about your problem. By "sentence complexity" do you mean "perplexity" (PPL)? For the "perplexity", we simply use GPT-2 to help us compute PPL. Can you tell me how you compute your "sentence complexity" and how you compare your results with us?