Improving Explainability In Personalized Systems By Extraction And Understanding