Design And Implement Of Dialogue Model Based On Emotion-enhanced

With the rapid development of the Internet and artificial intelligence technology,the dialogue model,has been widely applied in real life.Specifically,open-domain dialogue model which meets the need for companionship has shown huge potentials for research and application.Existing open-domain dialogue models are concerned with the fluency and content-related mostly instead of emotion for reply.Generating response with emotion has contributed to improve the experience of the interaction process and make artificial intelligence more user-friendly.Therefore,emotion-enhanced dialogue generation has huge significance for open-domain dialogue models.The purpose of emotional dialogue generation is to generate the response which is consistent with contextual emotion according to a given statement.There are many problems for the existing emotional dialogue generative models:uncontrollable emotion state and poor quality in generated responses.For the issues above,the paper focus on controllable emotional dialogue generation and emotional dialogue quality.The work and contribution are as follow:(1)For emotional dialogue quality,this paper proposes an emotional dialogue model based on generative adversarial network.The model consists of multiple generators and two discriminators.Each generator aims to generate response with specific emotion label and fraud the discriminator.two discriminators distinguish the content relevance and emotion type of the generated response respectively.Each generator updates its parameter by policy gradient.The model also uses a penalty-based objective function to promote diversity of response.Experiments show that the model has significantly improved the indicators of diversity and accurate emotional expression compared with previous studies.(2)For controllable emotional dialogue generation,this paper proposes an emotional dialogue model based on personalization.The keyword-emotion information is used as a key-value pair to establish a speaker database.Personal emotion information and input emotion information are combined according to the minimized emotional imbalance strategy to obtain external knowledge to guide and control the generation of emotions in responses.Experiments show that compared with the previous model,the model in this paper does not need to set a specific emotion category,and can generate smooth and diverse emotional responses.(3)Based on the previous model,this paper designs and implements an emotional dialogue system.The system includes three modules:client,server,and dialog generation.The functions and processes of each module are introduced in detail,and the usability of the system is proved through testing.Through the visual interface,dialogue models can provide good user experience when interacting with the user,and the feasibility of the model is also verified.