Cosmos QA: Machine Reading Comprehension with Contextual Commonsense Reasoning (EMNLP'2019)

This repository includes the source code and data for Cosmos QA.

Project Page and Leaderboard

• Please refer detailed examples, resources, as well as a leaderboard in our project page: Project Page
• To measure the performance of your system or create a submission to our leaderboard, please refer to Leaderboard for detailed instructions and policies.

Repo Details

• The training/dev/test datasets for Cosmos QA can be found in data/
• The BERT with multiway attention model can be ran and tuned with grid_search.sh. We also released the model file that we used in our paper for BERT multiway attention model Here

Requirements

Python3, Pytorch1.0, tqdm, boto3, requests