A novel feedback-based recommendation idea is introduced for service providers.
Reasons why people like or do not like something are discovered via semantic rules.
Factors that lead to a change in the opinions are analyzed.
Full processing pipeline to improve recommendation systems is proposed.