Implementation of 'Pretraining-Based Natural Language Generation for Text Summarization'

Paper: https://arxiv.org/pdf/1902.09243.pdf

Versions

• python 2.7

• PyTorch: 1.0.1.post2

Preparing package/dataset

0. Run: pip install -r requirements.txt to install required packages

1. Download chunk CNN/DailyMail data from: https://github.com/JafferWilson/Process-Data-of-CNN-DailyMail

2. Run: python news_data_reader.py to create pickle file that will be used in my data-loader

Running the model

For me, the model was too big for my GPU, so I used smaller parameters as following for debugging purpose. CUDA_VISIBLE_DEVICES=3 python main.py --cuda --batch_size=2 --hop 4 --hidden_dim 100

Note to reviewer:

• Although I implemented the core-part (2-step summary generation using BERT), I didn't have enough time to implement RL section.

• The 2nd decoder process is very time-consuming (since it needs to create BERT context vector for each timestamp).