But, we heard the same feedback from a few other users and we're thinking some ways to show how the model thinks and why it scores like this.

Thank you for trying and provide your valuable feedback.