See-Liu-Manning Text Summarization System

Context:
- it was developed by See et al. (2017).
- Resource(s): Software repository is available at https://github.com/abisee/pointer-generator
- System's Architecture:
  - It is based on Pointer-Generator Neural Network architecture.
- Training Systems and Tools:
  - Baseline models are training using the following tools:
    - Adagrad (Duchi et al., 2011) with learning rate 0.15 and an initial accumulator value of 0.1
    - Beam Search with beam size 4;
    - Gradient Clipping with a maximum gradient norm of 2;
    - See-Liu-Manning Coverage Loss Function.
Example(s):
- Training a model:
  python run_summarization.py --mode=train --data_path=/path/to/chunked/train_* --vocab_path=/path/to/vocab --log_root=/path/to/a/log/directory --exp_name=myexperiment
- Running a model on evaluation set:
  python run_summarization.py --mode=eval --data_path=/path/to/chunked/val_* --vocab_path=/path/to/vocab --log_root=/path/to/a/log/directory --exp_name=myexperiment
- Running beam search decoding:
  python run_summarization.py --mode=decode --data_path=/path/to/chunked/val_* --vocab_path=/path/to/vocab --log_root=/path/to/a/log/directory --exp_name=myexperiment
Counter-Examples:
See: Natural Language Generation System, Natural Language Processing System, Machine Translation System, ROUGE Metric, Meteor Universal Metric.

References