README.md

correlating-neural-and-symbolic-representations-of-language

Code repository for the paper:

Grzegorz Chrupała and Afra Alishahi. 2019. Correlating neural and symbolic representations of language. In Proceedings of ACL. https://arxiv.org/abs/1905.06401

Installation

Clone repo and set up and activate a virtual environment with python3

cd correlating-neural-and-symbolic-representations-of-language
virtualenv -p python3 .

Install pre-requisites annd the Python code (in development mode if you will be modifying something).

pip install -r requirements.txt
python setup.py develop

Download InferSent moodels and data and unpack them:

wget http://grzegorz.chrupala.me/data/infersent.tgz
tar zxvf infersent.tgz

Usage

• Train the LSTM models in experiments by running the run.py files.
• Run the main function in module rsa.report to generate the results:

python -c 'import rsa.report as R; R.main(open("report/results.tex", "w"))'