Research And Implementation Of Intellegent Voice Question Answer System For Program Design Course

With the advance of deep learning and speech recognition technology,intelligent voice question answering system has drawn more and more attention.For traditional question answering system,system usually get the answer of the question user inputted and return to the user by matching in the established rules of template or retrieving in a large information database,however the intelligent question answering system based on deep learning automatically generate answer for the user according to the question user inputted after learning from a large number of question and answer to the data set and extracting features.Considering the specific needs of the program design course,and based on the analysis of the domestic and foreign related technology of deep learning?speech recognition?question answering?micro service,this thesis studies and improves question answering system based on deep learning and speech recognition,puts forward the solution of intelligent voice question answering system and develop intelligent voice question answering system based on the Android platform.Firstly,on the basis of the relevant theoretical and technical research of intelligent voice question answering system,the functional requirements of the system are determined.Secondly,the author analyzes and studies the existing methods of speech recognition,decides to use the baidu speech recognition tool to recognize the voice user inputted.At the same time,human-computer interaction application is developed based on the open source Android smartphone operating system.User can ask question by voice,then the voice user inputted will be recognized into words by speech recognition module.Thirdly,the author analyzes and studies the existing implementation methods of micro service technology,decides to use micro service architecture to design the backend.High-performance distributed services framework Dubbo is used to connect different service,so that the service and the service can communicate each other.At the same time,use completely separated method of the frontend and backend to develop the corpus management system,the administrator can add,delete,update the corpus in the corpus management system.Finally,after analyzing the application of deep learning in question answering system and gradient descent algorithm,the author put forward the generating question and answer model,improve mini-batch gradient descent algorithm,take advantage of mini-batch gradient descent algorithm and improved mini-batch gradient descent algorithm to train the model with a large number of question and answer to the data set associated with program design.Experimental results show that corresponding reasonable answer generated by the model trained by improved mini-batch gradient descent algorithm according to question user inputted has higher accuracy.Besides,the author apply the speech recognition and generating question answer model to the intelligent voice question answering system and develop the intelligent voice question answering system based on android platform,then the system is used to provide question answer service for user.