A dialogue method and system based on generative sentiment memory and emotion state prediction
By employing a dialogue method based on generative emotional memory and emotional state prediction, the problem of insufficient emotional state modeling in dialogue systems is solved, thereby improving the consistency of dialogue context and the accuracy of responses. Furthermore, user emotional data is stored locally on the terminal device to protect privacy.
CN122451004APending Publication Date: 2026-07-24GUILIN UNIV OF ELECTRONIC TECH
View PDF 0 Cites 0 Cited by
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- GUILIN UNIV OF ELECTRONIC TECH
- Filing Date
- 2026-05-06
- Publication Date
- 2026-07-24
Smart Images
Figure FT_1 
Figure FT_2
Abstract
The application relates to the field of artificial intelligence dialogue technology and proposes a dialogue method and system based on generative sentiment memory and sentiment state prediction. The method converts user input into a structured sentiment state containing a sentiment category, a sentiment intensity and time information, and constructs a historical sentiment state sequence; when it is detected that the current sentiment state conflicts with the historical sentiment state and a preset trigger condition is met, a generative model is used to fuse the conflicting sentiment state, generate a unified sentiment representation and replace the historical sentiment state; the future sentiment state is predicted based on the historical sentiment state sequence; during dialogue generation, the historical sentiment state is weightedly searched based on the current sentiment state, the predicted sentiment state, the sentiment intensity and time decay, and the selected sentiment memory and the predicted sentiment state are jointly input into a dialogue generation model to generate a reply.
Need to check novelty before this filing date? Find Prior Art