Paragraph Ranker

This is a PyTorch implementation of a paper Ranking Paragraphs for Improving Answer Recall in Open-Domain Question Answering (Lee et al. 2018). Code implementations are based on the DrQA repository, and few additional experiments on entity based QA were conducted with @donghyeonk.

Requirements

• Install cuda-9.0
• Install Pytorch 0.4.1
• Python version >= 3.5 is required

Setups

Follow datasets and enviromental setups described in the DrQA repository.

Pretraining and Evaluation

# Pretrain reader model
$ python entityqa/reader/train.py

# Pretrain ranker model
$ python entityqa/ranker/train.py

# Evaluate QA pipeline
$ python entityqa/pipeline/predict.py --query-type SQuAD --ranker-type default --reader-type default

See codes for argument details of each file.