Text this: Learning Human Feedback from Large Language Models for Content Quality-aware Recommendation.