This document discusses a knowledge-based recommendation system that utilizes deep learning and sentiment analysis to assess users' mental conditions based on their posts on online social networks. The proposed method employs a Bidirectional Long Short-Term Memory (BLSTM) model and a sentiment metric called ESM2 to analyze user messages and recommend motivational content to alleviate stress. The work highlights the advantages of using deep learning over traditional algorithms like random forest and SVM for sentiment detection, emphasizing improved accuracy and personalized recommendations.