The key 'name'

[yjzhux]

DIDAN/didan_dataloader.py

Lines 52 to 54 in a36dc63

	for i, d in enumerate(self.real_arts):
	name = d['name']
	realarts2id[name] = i

Hi, what does the 'name' refer to? It should be a key stored in '*.bert.pt' which is the output of PreSumm. But there is no such a key in their code. See:
https://github.com/nlpyang/PreSumm/blob/70b810e0f06d179022958dd35c1a3385fe87f28c/src/prepro/data_builder.py#L320-L322

Another question: what are the hyper-parameters when you use PreSumm to extract language features? Thanks.

[rxtan2]

You can modify PreSumm to add in the name of the article when extracting the features.

what are the hyper-parameters when you use PreSumm to extract language features?

I am not sure about this. I don't recall setting any hyperparameters. All you have to do to get the language features is follow the steps in Option 2: process the data yourself.

[yjzhux]

Thanks. That's what I did.

I followed their steps and made the following changes for the hyper-parameters:

'-min_src_nsents', default=1
'-min_src_ntokens_per_sent', default=1
'-min_tgt_ntokens', default=0