Analysis of the Synergistic Role of User Natural Language Feedback and Behavioral Data in Adjusting Recommendation Strategies for Mental Health Platforms

This study examines the value of user natural language feedback in adjusting recommendation strategies for mental health applications and its synergistic effect with behavioral data. We collected 500,000 natural language feedback entries (including subjective experiences, emotional descriptions, and task reflections) alongside corresponding behavioral logs. A joint vector space was constructed integrating linguistic content features and behavioral characteristics, analyzing relationships among feedback types, emotional tendencies, and usage behaviors.In recommendation experiments, language feedback was integrated into strategy updates and compared against a baseline behavioral model (clicks, session duration, completion rate). Results showed that models incorporating language feedback achieved a 19.7% increase in user satisfaction metrics, a 22.4% rise in recommended content completion rate, and a 13.2% decrease in negative feedback rate.Further analysis revealed that users exhibiting negative emotional fluctuations relied more heavily on linguistic feedback to guide content selection, while user groups with weaker behavioral data demonstrated significant gains under this model. The study indicates that natural language feedback complements subjective information uncaptured by behavioral data, aiding in the development of more supportive mental health recommendation systems.