Advice on prediction and replacing the bert model

[wailoktam]

Hi, Thanks for your good work and I manage to run preprocessing and training on colab. Now I want to do two things:

1. predict without gold (generate a summary of text not found in training, validation, test set)
2. replace the bert model with that of a different language

Can you provide some general advice on these two tasks? It appear that all the translation, validation ,test do with a gold set. Thanks again.

[SebastianVeile]

Check this pull request for a Japanese model - https://github.com/nlpyang/PreSumm/pull/118