A user intention recognition method based on multimodal cross-improvement

Abstract

The present invention relates to the technical field of data processing and analysis, and specifically discloses a user intention recognition method based on multimodal cross-improvement, including constructing a user intention recognition model based on multimodal feature cross-improvement, and training the user intention recognition model with a privacy protection technology based on federal distillation, so that the user intention recognition result can be obtained through the user intention recognition model. The present invention fully explores the mutual influence relationship between different modal data representing users, and proposes to realize feature fusion of multimodal data for intent recognition based on a cross-improvement mechanism, and completes model training with a privacy protection technology based on federal distillation, which can not only effectively prevent data leakage and abuse, but also enhance user trust, ensure the legality and compliance of the research, and ultimately ensure the accuracy and efficiency of model training.

Images