Sentiment analysis and its explanation generation was done using Facebook's PyTorch-based model interpretability package called Captum. The dataset was collected from Canada Revenue Agency's twitter handle which sees regular taxpayer interaction.
An explanation of credit risk scoring model based on data from German Banks on 3000 borrowers
