iwangjian / plan4recdial Goto Github PK

Hi authors

Thank you very much for your interesting paper and the published source code.

Currently, we're trying to reproduce the reported results in the paper. However, the reproduced results are different with the reported ones. Specifically, we got the following logs after running and testing the TCP module.

Action ACC: 77.87%
Action Bi-ACC: 79.01%
Topic ACC: 35.45%
Topic Bi-ACC: 36.90%

Those results are much lower than your reported ones. I'm certain that I did nothing wrong since I strictly followed your instructions in this repo. Looking forward to your response.

Hi authors. Could you provide your implementation of the KERS model ?

Amazing work!
I am curious if your model also supports English dataset or not.

If yes, could you please provide me a sample dataset in English..?

Thanks!