search

iwangjian / textsum-gan

Tensorflow re-implementation of GAN for text summarization
134 stars 56 forks source link

parser.add_argument('--decode_dir', required=True, help="root of the decoded directory"). #25

Open istiakahmad opened 3 years ago

istiakahmad commented 3 years ago

In gen_sample.py file, parser.add_argument('--decode_dir', required=True, help="root of the decoded directory") this line refers to the decoded directory with two child directories (reference and decoded). I don't understand how to create these directories effectively. Please help me.

I run the gen_sample.py file and got this output: python3 gen_sample.py --data_dir data/ --decode_dir decode_dir --vocab_path data/vocab

vocab length: 199869 positive samples: 0 negative samples: 0 file saved: data/discriminator_train_data.npz

iwangjian commented 3 years ago

I'm sorry that this repo released 3 years ago is confusing. The ${decode_dir}/reference denotes a dir for positive samples, which can be ground-truth summaries for the given documents, the ${decode_dir}/decoded denotes a dir for negative samples, which can be randomly sampled summaries by negative sampling or decoded summaries by a coarse summarization model.

istiakahmad commented 3 years ago

Thanks for the guidelines.

istiakahmad commented 3 years ago

Sorry, i can't generate positive and negative sample. Could you please help me to

  1. How can I generate positive and negative sample?
  2. What will the extension or format of positive and negative sample file?