Use of BertExt model

[rudra0713]

Hi,
If I wish to find the most important sentences using the BertExt model prior to generating summaries, what's the best way to do it? For example, can I use the CNN/DM Extractive model under the trained model section from the Presumm work (https://github.com/nlpyang/PreSumm/tree/dev)?

[rudra0713]

@maszhongming I checked the Presumm model's github page and they have released several trained models. For your implementation, did you use one of these models or did you train a model from scratch?

Thanks in advance.