Research And Application Of Conversational Emotion Recognition Based On Attention Knowledge Enhancement

Conversational sentiment recognition is an important component of sentiment recognition tasks.It has received continuous attention in areas such as natural language processing and text mining.The task of conversational sentiment recognition aims to capture the sentiment dynamics of the user.It has important application prospects in dialogue systems,opinion mining,legal trials,interviews,and electronic medical services.With the popularity of social media,more and more users express their views online.Users often rely on some common-sense knowledge to express their opinions and the conversation has some reasons for the speaker to generate a certain emotion.The model lacks common-sense cognitive and inferential skills.These problems lead to inaccurate recognition of conversational emotions.This paper proposes an attention knowledge enhancement model for conversational sentiment recognition.The model aims to use attention mechanisms to integrate knowledge into conversations and enrich the semantic information in conversations.And it uses multi-round reasoning to extract conversational contextual features for final sentiment classification.The main research work in this paper is as follows.(1)To address the problem that conversational emotion recognition models lack the ability of common sense cognitive,this paper proposes an incremental encoding model based on attentional knowledge enhancement.The model introduces a common-sense knowledge and sentiment lexicon and models external knowledge with conversational content.Firstly,the model uses a double-layer emotion graph attention mechanism to enrich the semantics of the conversation.The way uncovers the hidden emotional information in the conversation.Secondly,the model uses an incremental encoding module to encode the conversational context and extracts the conversation-level context and the utterance-level context respectively.It further extracts the key contexts of the conversation.Finally,the experimental results show that the model has improved accuracy compared to other models.The improvement is 1.17%,3.76%,1.28%,1.07% and 1.25% across the five datasets respectively.And the results demonstrate the effectiveness of common-sense knowledge for conversational emotion recognition tasks.(2)To address the problem that the conversational emotion recognition model lacks certain inference capability,this paper proposes a multi-round reasoning model based on attentional knowledge enhancement.The model uses external knowledge to enrich the semantics of the conversation and employs a multi-round reasoning process to integrate the affective cues in the conversation.Firstly,the multi-round reasoning module obtains the global context as reference information for the conversational sentiment.Secondly,the model integrates its contextual cues in multiple rounds and uses a multi-task learning framework to reduce model loss.Finally,the experimental results show that the model integrates common sense knowledge and reasoning processes to improve the accuracy of model processing.The improvement is 1.04%,1.26% and 2.76% on the three datasets respectively.The results demonstrate the effectiveness of the multi-round reasoning model based on attentional knowledge enhancement.(3)In order to verify the practical application of the multi-round reasoning model based on attentional knowledge enhancement,the paper designs and develops a conversational emotion recognition system based on attentional knowledge enhancement.The system achieves dynamic capture of psychological states and focus on user’s mental health.The experimental results show the effectiveness of an attentional knowledge enhancement-based conversational sentiment analysis model for practical problems.