Skip to content
Goal driven language generation using knowledge graph A2C agents
Python
Branch: master
Clone or download
Latest commit 1c99190 Jan 14, 2020
Permalink
Type Name Latest commit message Commit time
Failed to load latest commit information.
data Initial Commit. Dec 30, 2019
kga2c KG cleanup Jan 14, 2020
.gitignore KG cleanup Jan 14, 2020
LICENSE.md Create LICENSE.md Dec 27, 2019
README.md KG cleanup Jan 14, 2020

README.md

KG-A2C

Goal driven language generation using knowledge graph A2C agents. This code is accompanies the paper Graph Constrained Reinforcement Learning for Natural Language Action Spaces.

Bibtex

@inproceedings{
ammanabrolu2020graph,
title={Graph Constrained Reinforcement Learning for Natural Language Action Spaces},
author={Prithviraj Ammanabrolu and Matthew Hausknecht},
booktitle={International Conference on Learning Representations},
year={2020},
url={https://openreview.net/forum?id=B1x6w0EtwH}
}

Quickstart

Install Dependencies: Jericho, Redis, Pytorch >= 1.2

pip3 install --user jericho
pip3 install torch torchvision
sudo apt-get install redis-server

Download and extract Stanford CoreNLP then start the OpenIE server:

cd stanford-corenlp-full-2018-10-05/ && java -mx8g -cp "*" edu.stanford.nlp.pipeline.StanfordCoreNLPServer -port 9000 -timeout 15000

Train KG-A2C

cd kga2c && python train.py --rom_file_path path_to_your_rom --openie_path path_to_your_openie_install --tsv_file ../data/rom_name_here
You can’t perform that action at this time.