CORGI

This is the code for the Conversational Neuro-Symbolic Commonsense Reasoning.

Dependencies

1. PyTorch
2. Spacy
3. higgingface en_coref_lg model: https://github.com/huggingface/neuralcoref-models/releases/download/en_coref_lg-3.0.0/en_coref_lg-3.0.0.tar.gz
4. python2
5. numpy
6. py
7. rpython

Navigation

1. logging/ folder to store user interaction logs under <USER_NAME> exaplained below.
2. testnet/ contains the code for our neuro-symbolic theorem prover built on top of spyrolog
3. data/ contains our proposed commonsense reasoning benchmark. Look at data/REAME.md for a data description.
4. net.py, attention.py, config.py, modeltest.py are files relevant to the inference for our neuro-symbolic theorem prover
5. TypeDict.json modified dictionary of types built on top of Aristo tuple KB v1.03 Mar 2017 Release
6. prolog_info-reasoning-17.pkl, model_testing-reasoning-17.tch pretrained models for the neuro-symbolic theorem prover
7. facts7.txt, functor_arity7.txt knowledge base of commonsense facts
8. user-study-data.txt statements used in the user study

Running the code:

in order to run the code use prompt:

python2 reasoning.py --user <USER_NAME> --engine /testnet

additional arguments are:

• --verbose true (can be used for debugging purposes to see more printouts)
• --resume <STATEMENT_NUMBER> (can be used to resume the study at a certain statement if needed)