Research On Key Technology Of Task-oriented Multi-round Emotional Dialogue

Multi-round dialogue and emotion detection technologies are important research directions in natural language processing(NLP).With the increasing needs of users and the continuous maturity of neural network,more and more task-based dialogue processes introduce"artificial intelligence assistant" to replace manual reply,and use intelligent interactive system to solve the specified needs of users in specific task areas,which can not only save personnel resources,but also solve the needs of users in time.With the continuous development and application of multi round technology,researchers gradually began to pay attention to the detection and analysis of users’ emotion in the process of dialogue,so as to meet users’ emotional needs,so as to better complete the specified task.At present,multiple rounds of emotional dialogue tasks are mainly studied within the scope of open domain dialogue system,in order to understand users’ emotion in the process of chatting and have emotional dialogue with users.In the process of task-oriented dialogue,the system also needs to understand the user’s emotion.Different from the open domain,in the process of task-oriented dialogue,understanding emotional information is to help users better complete the target task.The main reason is that when the user’s emotion is in a negative state such as angry,sad,etc,statements with emotional comfort information can be added in the reply process,so as to alleviate the user’s emotion and better complete the dialogue task.When studying the key technologies of Task-based multi round emotional dialogue,this paper focuses on the emotion detection in the context of task-oriented multi round dialogue,discusses how to comprehensively and accurately detect users’ emotion in the mode of APP interaction,and discusses all possible input types through the investigation of related technologies of multi-modal emotion detection.The nested long-term and short-term memory network is combined with multimodal emotion detection technology to comprehensively detect the possible emotional input of users and improve the accuracy of emotion detection.Based on the research of emotion dictionary and emotion classification algorithm,an emotion detection model based on dictionary and emotion classification algorithm is proposed to comprehensively detect users’ emotion input and improve the accuracy of emotion detection.The innovation of this paper is mainly as follows:In view of the actual use scenario of task-based multi-round dialogue,the experimental data set mainly composed of negative emotion labels is selected and constructed.According to the experimental needs,the experimental data are manually labeled to make the experiment more suitable for the actual scene and provide data guarantee for the smooth progress of the experiment.At the same time,an emotion detection technology based on nested long-term and short-term memory network is proposed.The nested long-term and short-term memory network deepens the number of long-term and short-term memory network layers,increases the internal and external network levels,and solves the problem of weak ability to retain longer-term memory in the original long-term and short-term memory network.At the same time,based on the research on the multi-modal emotion detection technology of nested long-term and short-term memory network,the context-based nested long-term and short-term memory network structure is built for the application scenario of multi-round dialogue emotion detection,and the multi-modal input of users in the dialogue process is comprehensively detected.A human-machine dialogue system for emotion recognition of Winter Olympics service mix is designed and implemented,and the proposed method is verified.The multi-round dialogue module includes natural language understanding module,dialogue state tracking module,dialogue strategy management module and natural language generation module.Through the use of "Bert+rules",an assembly line pipeline type multi round dialogue system is built,and a data set containing Winter Olympic events,time,venues and other information is created in the process.Then,based on the surge of people’s negative emotional state under the background of the epidemic,this paper analyzes the emotional crisis in reality,and proposes to establish an emotional early warning module in the process of task dialogue to detect and warn the possible dangerous tendencies of users in time,so as to reduce the occurrence of possible tragedies.At the same time,this paper studies and analyzes the psychological construction process of colleges and universities that this module can be applied.Finally,the engineering implementation and use of Task-based multi round emotional dialogue system are studied,and the application scenario and functional structure of the system are explained through demand analysis.Then the overall structure of the system and the details in the architecture are analyzed.The interaction mode and specific process are described in detail,and the performance test is carried out to verify its practicability.