Research And Application Of The Optimization Strategy Of Speech Interaction Error Feedback Experience

Speech interaction is one of the human-computer interaction modes,which has the advantages of being more natural and easy to learn than other human-computer interaction modes.Although the test results in the laboratory show that speech recognition has reached an extremely high level,the accuracy of automatic speech recognition will decline rapidly as the applicating scenarios of speech interaction technology continue to be complicated.In view of the fact that most of the current speech interaction error feedback is repetitive,monotonous,and lack of emotion,causing users' dissatisfaction with the speech interaction intelligent assistant.And the current research on this problem is extremely scarce,so choosing the speech interaction error feedback as the starting point of this article.Based on the analysis of human behavior when dealing with contradictions and economics theory,proposing three hypotheses in combination with emoticons,nonverbal communication,and diminishing marginal utility.Through three series of experiments to find out what emojis have a soothing effect and probe into weather the role of emojis in speech interaction error feedback,and studied the feedback mechanism of single error feedback and continuous error feedback in order to provide a reference for the optimization design of speech interaction error feedback experience.Experiment one uses PPT to simulate the error feedback of speech interaction,and explores what emojis that has a soothing effect,and the influence of its application in the error feedback of speech interaction on the emotional valence and the degree of personification.The results show that the emoticons corresponding to the soothing smile,blushing,and embarrassing emoticons are "grin","shy",and "facepalm";speech interaction error feedback corpus with emoticons can significantly improve user experience in the error feedback of speech interaction;the type of apology feedback is best combined with "shy";the type of humorous feedback is better combined with "facepalm",and apology corpus is the first choice for speech interaction error feedback.The simulation method of experiment two is the same as experiment one.Investigated the influence of repetition of current speech interaction error feedback corpus on user satisfaction(valency).The results show that the strategy of current speech interaction error feedback will cause the phenomenon of diminishing marginal utility;user's interview results show that users generally can only accept three errors.On the basis of the previous experimental conclusions,the experimental corpus of speech interaction error feedback in experience three was determined to be a random combination of apology,"shy" + apology,and "facepalm" + humor.The simulation method of the experiment is the same as that of Experiment one,we explore the difference in satisfaction and personification degree of different error feedback corpus combinations in experience three.The results show that users think that the continuous error feedback of apology,"shy" +apology,"facepalm" + humor is the best,that is,the continuous error feedback corpus strategy with logic and increasing utility is better.Then,based on the experimental conclusions,proposing a speech interaction error feedback experience optimization design strategy,and applies it into practice for evaluation.The results show that the optimized intelligent assistant can significantly improve user's satisfaction and perceptual personification.However,the number of consecutive error feedbacks is greater than three,which will have certain limitations.Finally,a general summary of the full text is made,and the deficiencies of this research and the direction of research in the future are pointed out.