/Chatbot-Common-Sense-Evaluation

Evaluating the Sensibility of Generated Responses by using existing Common Sense Knowledge Graphs

Primary LanguageJupyter Notebook

Chatbot-Common-Sense-Evaluation

Evaluating the Sensibility of Generated Responses by using existing Common Sense Knowledge Graphs

Local Setup

1. This app uses COMET. Download COMET files into the current directory and follow the setup instructions provided.
https://github.com/atcbosselut/comet-commonsense
2. To run experiments with DiSCoL, follow the steps to start the discol server and update the URL in discol_dataset_generation.py.
https://github.com/PlusLabNLP/Dialogue_System_Hackathon

Pre-trained model can be downloaded at: https://drive.google.com/file/d/1PtWQbYJfkZPI_Uql94ALM0HRF2J2CXX5/view?usp=sharing
Generated non-commonsense samples can be found under results.