A digital human multi-dimensional semantic sentiment analysis system and method

By employing a multi-dimensional semantic sentiment analysis method and a BERT-CSAM hybrid model architecture, the problems of insufficient granularity and high annotation costs in sentiment analysis during digital human interaction are solved, achieving efficient and accurate sentiment analysis and interactive feedback, thereby improving the emotional expression and interactive experience of digital humans.

CN122113937APending Publication Date: 2026-05-29SUZHOU GOLD MANTIS EXHIBITION DESIGN ENG
View PDF 0 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN ยท China
Patent Type
Applications(China)
Current Assignee / Owner
SUZHOU GOLD MANTIS EXHIBITION DESIGN ENG
Filing Date
2026-01-15
Publication Date
2026-05-29

AI Technical Summary

Technical Problem

Existing sentiment analysis technologies suffer from insufficient analytical granularity, difficulty in decoupling semantics and sentiment in digital human interaction scenarios, and high costs in constructing high-quality labeled datasets, making it difficult to achieve multi-dimensional, fine-grained sentiment analysis and efficient data construction.

Method used

We employ a multi-dimensional semantic sentiment analysis approach, construct a training dataset through a semi-supervised annotation process, train the model using the BERT-CSAM hybrid model architecture, and combine it with an adaptive learning rate optimization algorithm to achieve multi-dimensional sentiment feature extraction and accurate interpretation, thereby reducing computational resource consumption and improving model efficiency.

Benefits of technology

It enables multi-dimensional quantification and precise interpretation of textual sentiment, improves the accuracy and subtlety of sentiment analysis, enhances the interactive affinity and immersion of digital humans, and reduces annotation costs and computational resource consumption.

โœฆ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN122113937A_ABST
    Figure CN122113937A_ABST
Patent Text Reader

Abstract

The application discloses a digital human multi-dimensional semantic sentiment analysis system and method, relates to the technical field of digital humans, and through multi-dimensional sentiment quantification, training data sets for neural network model training are first marked with multi-dimensional semantic sentiment, and then are used for training of the neural network model; compared with a conventional single-sentiment-label classification traditional sentiment analysis mode, the neural network model and the digital human applying the model can recognize more dimensional semantic sentiment information, semantic sentiment analysis is changed from conventional qualitative analysis to quantitative analysis, and can capture mixed, complex and subtle emotional states, so that the sentiment analysis precision and delicacy are significantly improved.
Need to check novelty before this filing date? Find Prior Art