This is a rule-based system with learned rules using rule extraction method to extract rules from deep neural network. The task is whether this person can get the loan from the bank or not.
- Load the program using swipl
- run the Learned.pl and load the example Sceanrio1_dialogue.pl
- type 'chat' to start the conversations
Note: The parser.pl is to format the text rule into the prolog. The rules are extracted from the REM method. We've changed the dataset into financial and trained the model, then extract the rules from the trained model. It has conflict rules in the raw rules, because machine learning model can't be 100% accurate. In the user evaluation, we removed the conflict rule by ourselves in order to avoid the quality of our explanation.