Research On Human-Computer Interaction Topic Recommendation Method Based On Knowledge Graph

With the continuous development of intelligent dialogue systems,the way of human-computer interaction has undergone tremendous changes.The original human-computer interaction based on command to feedback has been gradually broken,and human-computer interaction is developing in a more convenient,natural,and intelligent direction.The machine is developing from passively receiving user information to actively understanding the user's intention.With the continuous accumulation of user data,big data and artificial intelligence technology make robots understand "self" better than users in some scenarios,but in the current intelligent voice dialogue system,topic development has been driven by users.This can easily cause problems,such as missing points of interest,decreased interest,and termination of conversations.Therefore,this thesis focuses on the problems in the intelligent speech system,starting from two aspects: how to choose the appropriate topic in the dialogue process and how to guide the topic.The main contents are as follows:In the process of recommending human-computer dialogue topics,the user's dialogue content is short and the topic is time-sensitive.This thesis proposes a topic recommendation algorithm based on knowledge graph.The algorithm first selects candidate topics for users based on the characteristics of similar users with the same hobbies.The topic prediction of human-computer dialogue is transformed into the problem of user's click probability on candidate topics.Secondly,the knowledge graph is used to enhance the feature representation of the user's expressed content.Finally,use the attention neural network model to predict candidate topics.Experiments show that this algorithm is significantly higher than other models in terms of accuracy and precision in the recommendation of human-computer dialogue topics.Aiming at the problem of how to guide topics for users in the process of human-computer dialogue,this thesis proposes a game-based topic guidance model for human-computer interaction.This model proposes human-computer interaction emotional friendliness and human-computer interaction content friendliness and its quantitative representation method according to the human-human interaction relationship,which is used to analyze and quantify the human-computer interaction relationship representation.By analyzing the game relationship of topics between people,the topic game process of human-computer interaction is modeled to determine whether to guide topics and which topics to guide.Through the above methods to achieve topic guidance and reply to the user.Experiments show that this model can guide topics without destroying the harmony between human and machine.After the guidance,the user's emotion and interest will be greatly improved,and the satisfaction of human-computer interaction will be improved.