(from Interactive Task Learning for Social Robots: a Pilot Study in IEEE IROS 2023)

FurhatDriver contains the client skill that must be run on the furhat robot. This skill is also responseible for hosting the website. Compile and run per the Furhat Skill developer docs: (https://docs.furhat.io/skills/)

social_itl contains the server code that performs all of the interaction logic. To setup:

1. Install dependencies from requirements.txt
2. Run dataset.py to generate the synthetic dataset
3. Run train_anonymizer.py to train the masking BERT model
4. Run train_parser.py to train the parsing model
5. Run sentence_classifier.py to train the sentence classifier

To run the full system, start the furhat driver and run python -m social_itl --host <ip address> with the IP address of the furhat robot.