Mining affective text to improve social media item recommendation

•We propose a sentiment-aware social media recommendation framework.•An ensemble learning-based method is proposed to classify sentiments from affective texts.•We conduct comprehensive experiments to verify the effectiveness of the proposed methods.