omprasad93 / hmnet-end-to-end-abstractive-summarization-for-meetings Goto Github PK

@article{zhu2020end,
  title={End-to-End Abstractive Summarization for Meetings},
  author={Zhu, Chenguang and Xu, Ruochen and Zeng, Michael and Huang, Xuedong},
  journal={arXiv preprint arXiv:2004.02016},
  year={2020}
}

1. I used CNN in Transformer-PositionwiseFeedForward.
2. Role vectors, pos tags, and named entity tags are not used. 
(When using a role vector, performance was lower than not utilizing role vector. So please do not hesitate to advise me about this.)

python main.py --mode train --save_path path_to_save_the_model

python main.py --mode eval --model_path trained_model_path --gen_max_length 500