Research On Human-Computer Interactive Emotional Anthropomorphic Strategies

As an important way to reflect the application value of artificial intelligence technology,human-computer interaction system has attracted the attention of many scholars in various fields,and its products have gradually entered people's daily lives.Affected by the great satisfaction of material needs,people have begun to desire and pursue the emotional fit brought by human-computer interaction.It is hoped that robots can meet the needs of daily interaction while having the ability of cognitive affective computing to generate advanced anthropomorphic emotions.In order to enhance the cognitive affective computing capabilities of robots,this thesis focuses on the problems existing in the research of robots' cognitive emotions,and conducts research work from two aspects: open and closed domains.The main research contents are as follows:Aiming at the problem of rough emotion and incoherent expression of robot answer responses in existing open-domain human-computer interaction systems,a cognitive affective interaction model of robot based on reinforcement learning combining instant feedback and long-term trends is proposed according to PAD emotion space.Firstly,the process of human emotion generation is simulated by social psychology to extract emotion features;the process of human emotion generation is simulated based on social psychology to extract emotion features.Secondly,the global overall planning characteristics of reinforcement learning is used to establish the relationship between the response emotional state and the context emotional state,thereby modeling the robot emotion generation process;Then,the emotional features are incorporated into the model reward mechanism to evaluate the interactive emotional state,so as to realize the model updating and the selection of the optimal emotional strategy,according to which the transition probability of the robot's current emotional state is updated.Then,the emotional features are incorporated into the model reward mechanism to evaluate the interactive emotional state,so as to realize the model updating and the selection of the optimal emotional strategy,according to which the transition probability of the robot's current emotional state is updated.Finally,the optimal response emotion value of the robot in the continuous emotion space is calculated by combining the 6 basic emotion values and the updated transition probability.The experimental results show that the proposed model not only effectively increases the exquisiteness,continuity and enthusiasm of the robot's emotional expression,but also effectively enhances the user's willingness to interact in the process of human-computer interaction.Aiming at the problem of semantic accuracy of robot answer responses and less attention to the rationality of robot emotions in existing closed-domain human-computer interaction systems,a cognitive affective interaction model of customer service robot based on Attention mechanism and Long Short-Term Memory network is proposed according to PAD emotion space.First,the Attention mechanism and the Long ShortTerm Memory network are used to learn multiple rounds of historical emotional interaction processes to realize the autonomous prediction of the robot's optimal affective strategy.Secondly,the candidate answer set is initially selected based on the semantic information of the user's current interactive input content,and the answer with the highest confidence ranking is selected as the candidate answer.Finally,the semantic similarity and emotional matching of candidate answers are comprehensively considered to select the optimal conversation strategy to respond.The experimental results show that,compared with the comparison model,the proposed model considers both the semantic accuracy of the reply content and the degree of emotional matching,which not only effectively increases the effectiveness of customer service robot information retrieval,but also effectively improves users' satisfaction with customer service robot services.