A user emotion recognition method based on text information, medium and device
By using text-based user intent recognition and emotion contagion models, combined with a spatial distance weighting function, the problem of inaccurate emotion recognition in digital customer service systems when voice and video input are lacking is solved, achieving highly accurate emotion assessment and service response in privacy-preserving scenarios.
Patent Information
- Application Number
- CN202610077627.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2026-01-21
- Publication Date
- 2026-09-11
- Estimated Expiration
- 2046-01-21
AI Technical Summary
Existing digital customer service systems, lacking voice and video input, rely solely on text content for emotion analysis, resulting in low accuracy and an inability to effectively capture users' true emotional states. This is especially problematic in high-density crowds where they cannot distinguish between shared and unshared anxieties, leading to mismatched service responses and a reduced user experience.
By using text-based user intent recognition and emotion contagion models, combined with spatial distance weighting functions, the influence of individual and group emotions is calculated. A two-stage emotion assessment mechanism is designed to distinguish the contagion weights of common and non-common anxieties, thus compensating for the shortcomings of emotion recognition under a single text modality.
In privacy-protected scenarios, it improves the accuracy of emotion recognition and the quality of service response, and can accurately compensate for users' emotional assessment when multimodal input is lacking. It significantly enhances the emotional perception ability of digital customer service, especially in providing timely and effective reassurance and solutions in the event of emergencies.