Dialogue emotion recognition method combining character clues and emotion uncertainty

By combining personality cues and emotional uncertainty modeling, pre-trained model and personality detection model are used to construct a dialogue emotion recognition framework, which solves the problem of personality traits and emotional dynamics neglect in traditional methods, and improves the accuracy of dialogue emotion recognition.

CN120277215APending Publication Date: 2025-07-08TAIYUAN UNIVERSITY OF TECHNOLOGY
View PDF 6 Cites 0 Cited by

Patent Information

Application Number
CN202510338153.5
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2025-03-21
Publication Date
2025-07-08

AI Technical Summary

Technical Problem

The prior art ignores the speaker's personality traits in dialogue emotions recognition, which makes it difficult to effectively capture emotional dynamics and uncertainties. Traditional methods have limitations in digging deep emotional dynamics clues.

Method used

The pre-trained model is used for fine-tuning, combining the bidirectional gating recurrent unit and personality detection model, through personality cues and emotional uncertainty modeling, and using the probability distribution encoder and Markov chain Monte Carlo sampling method, a dialogue emotion recognition framework is constructed to achieve a fusion representation of discourse-level and global personality traits.

Benefits of technology

It improves the accuracy of dialogue emotions recognition, can identify human emotions more accurately, solves the problem of ignoring personality traits and emotional uncertainty in traditional methods, and provides a more comprehensive background representation.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN120277215A_ABST
    Figure CN120277215A_ABST
Patent Text Reader

Abstract

The invention provides a dialogue emotion recognition method combining character clues and emotion uncertainty, and belongs to the technical field of natural language processing. The problem of incomplete utterance representation in current text dialogue emotion recognition is solved; comprising the following steps: constructing utterance-level background representation; on the basis of the utterance-level background representation, constructing a dialogue-level background representation; obtaining a character representation vector of each utterance in the dialogue, and obtaining a character feature vector of each speaker by adopting a statistical model; obtaining a character perception vector; converting the dialogue-level background representation into multivariable Gaussian distribution by adopting a probability distribution encoder; sampling a text time sequence background vector from multivariable Gaussian distribution; the text time sequence background vector and the character perception vector are converted into the same concept space, and then the text time sequence background vector and the character perception vector form a new vector through splicing operation; predicting an emotion category of each utterance by adopting an emotion classifier; the method is applied to dialogue emotion recognition.
Need to check novelty before this filing date? Find Prior Art

Citation Information

Patent Citations

  • Text emotional tendency determination method and system

    CN105095183A

  • User emotion analysis method and system integrating personality characteristics and social relations of users

    CN113806476A

  • Personality-enhanced dialogue emotion prediction model and construction method thereof

    CN114970561A

  • Emotion recognition method and related device

    CN118862901A

  • Emotion recognition method and system based on dialogue situation knowledge base and comparative learning

    CN119202854A