Looking for breakthrough ideas for innovation challenges? Try Patsnap Eureka!

Dialogue emotion recognition network model based on double knowledge interaction and multi-task learning, construction method, electronic equipment and storage medium

A multi-task learning and emotion recognition technology, applied in the field of natural language processing, can solve problems such as limited, weak models, ignoring the direct interaction between words and knowledge

Active Publication Date: 2021-10-22
HARBIN INST OF TECH
View PDF9 Cites 5 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0007] In view of this, this application provides a dialogue emotion recognition network model, construction method, equipment and storage medium based on dual knowledge interaction and multi-task learning to solve the problem that the existing ERC model ignores the direct interaction between discourse and knowledge; the models all use Auxiliary tasks that are weakly related to the main task, can only provide limited emotional information for ERC tasks

Method used

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
View more

Image

Smart Image Click on the blue labels to locate them in the text.
Viewing Examples
Smart Image
  • Dialogue emotion recognition network model based on double knowledge interaction and multi-task learning, construction method, electronic equipment and storage medium
  • Dialogue emotion recognition network model based on double knowledge interaction and multi-task learning, construction method, electronic equipment and storage medium
  • Dialogue emotion recognition network model based on double knowledge interaction and multi-task learning, construction method, electronic equipment and storage medium

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0086] Embodiment 1 of the present application provides a dialogue emotion recognition network model based on dual knowledge interaction and multi-task learning (see figure 1 ), including: a task definition module, which is used for a set of dialogues in a given dialogue data set, and predicts the emotional label of each target utterance given the dialogue history information; the encoder uses the XLNet encoder to model the dialogue history Information; knowledge integration module, used to extract common sense knowledge, and obtain knowledge enhancement representation based on graph attention network; self-matching module, used for discourse-knowledge interaction; dialogue emotion recognition module, combined with dialogue history information to predict the current target discourse The emotion label of ; the emotional polarity strength prediction task module is used to introduce the knowledge strongly related to the main task into the model, and utilize the setting of multi-ta...

Embodiment 2

[0088] Embodiment 2 of the present application provides a method for constructing a dialogue emotion recognition network model based on dual knowledge interaction and multi-task learning (see figure 2 ), the method is specifically:

[0089] Step 1: Given a collection of dialogues in the dialogue data set, predict the emotional label of each target utterance given the dialogue history information;

[0090] The dialogue emotion recognition task is defined as follows: Given where i=1,...,N, j=1,...,N i , represents a set of dialogue pairs {utterance, label} in the dialogue dataset. Dialogue X contains N utterances, each utterance X i contains N i words, expressed as every X i by p(X i ) ∈ P, where P is the set of speakers. Discrete value Y i ∈S is used to represent sentiment labels, where S represents the set of sentiment labels, and |S|=h c . The goal of the dialogue emotion recognition task is to predict each target utterance X given the dialogue history informati...

Embodiment 3

[0143] Embodiment 3 of the present application provides an electronic device, see image 3 , an electronic device in the form of a general-purpose computing device. Components of an electronic device may include, but are not limited to: one or more processors or processing units, memory for storing computer programs that can run on the processors, connections to different system components (including memory, one or more processors or processing unit) bus.

[0144] Wherein, when the one or more processors or processing units are used to run the computer program, execute the steps of the method described in the second embodiment. The types of processors used include central processing units, general purpose processors, digital signal processors, application specific integrated circuits, field programmable gate arrays or other programmable logic devices, transistor logic devices, hardware components or any combination thereof.

[0145] Wherein, the bus refers to one or more of ...

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More

PUM

No PUM Login to View More

Abstract

The invention discloses a dialogue emotion recognition network model based on double knowledge interaction and multi-task learning, a construction method, electronic equipment and a storage medium, and belongs to the technical field of natural language processing. The problem that an existing Emotion Recovery in Conversion (ERC) model neglects the direct interaction between the utterance and the knowledge is solved. And when an auxiliary task weakly related to the main task is used, limited emotion information can only be provided for the ERC task. According to the method, common knowledge in a large-scale knowledge graph is utilized to enhance word level representation. A self-matching module is used to integrate knowledge representations and utterance representations, allowing complex interaction between the two representations. And a phrase-level sentiment polarity intensity prediction task is taken as an auxiliary task. A label of the auxiliary task comes from an emotion polarity intensity value of an emotion dictionary and is obviously highly related to an ERC task, and direct guidance information is provided for emotion perception of a target utterance.

Description

technical field [0001] The present application relates to a dialogue emotion recognition network model, construction method, electronic equipment and storage medium, in particular to a dialogue emotion recognition network model, construction method, electronic equipment and storage medium based on dual knowledge interaction and emotional polarity intensity perception multi-task learning The storage medium belongs to the technical field of natural language processing. Background technique [0002] Conversational emotion recognition has attracted much attention in the field of natural language processing in recent years due to the explosion of publicly available dialogue data. Dialogue emotion recognition aims to identify the emotion of each utterance in a dialogue. This task requires the machine to understand the way emotions are expressed in the dialogue. Since the ERC model enables machines to understand the emotions in human conversations, which in turn enables machines t...

Claims

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More

Application Information

Patent Timeline
no application Login to View More
IPC IPC(8): G06F16/35G06F16/36G06N5/02G06N3/08G06F40/284G06F40/242
CPCG06F16/35G06F16/367G06N5/02G06N3/08G06F40/284G06F40/242
Inventor 孙承杰解云鹤刘秉权季振洲刘远超单丽莉林磊
Owner HARBIN INST OF TECH
Who we serve
  • R&D Engineer
  • R&D Manager
  • IP Professional
Why Patsnap Eureka
  • Industry Leading Data Capabilities
  • Powerful AI technology
  • Patent DNA Extraction
Social media
Patsnap Eureka Blog
Learn More
PatSnap group products

Browse by: Latest US Patents, China's latest patents, Technical Efficacy Thesaurus, Application Domain, Technology Topic, Popular Technical Reports.

© 2024 PatSnap. All rights reserved.Legal|Privacy policy|Modern Slavery Act Transparency Statement|Sitemap|About US| Contact US: help@patsnap.com