Design And Implementation Of Task-Oriented Conversation Application Framework

With the development of science and technology,chat robots are no longer just the existence of science fiction movies,and the conversation system has become a rising concern because of its huge potential and market value.The task-oriented conversation application framework could help people accomplish specific tasks and reduce labor costs.In traditional task-oriented dialogue application development,application developers should not only pay attention to specific application scenarios,but also should complete the common functions of task-oriented conversation applications such as intent classification and entity recognition,dialogue management and so on.Introducing the task-oriented dialog application framework to provide the versatility and basic functionality of the task-oriented dialogue application can greatly reduce the complexity and workload.The task-oriented conversation application framework is designed and implemented in this thesis.This framework provides a Web-based interface,which allows users to input text content,analyzes the inputs to identify users'intentions.When it is necessary,multi-turn conversation is needed,then system invokes the corresponding service interfaces to users'intentions to meet the users' needs and reply users using Web-based interfaces.In this thesis,the joint intention judgment model based on rules and neural networks is used to identify the intent of user's input.The rule-based model is implemented by AIML rule model.The neural network intent recognition module uses LSTM to obtain the semantic features and output the probability of each feature,then use the SoftMax function to normalize the classification output.As a result,the joint model is easy to make modifications and intelligently trainable.This thesis first introduces the research background and related technologies of the task-oriented conversation application framework.And based on the investigation of current researches and applications on the conversation application,the requirements of the task-oriented conversation application framework are analyzed.Then the architecture design and the design of each module are specified,as well as the implementation.In the end,the framework was tested and a sample application was developed based on this framework,which proved the validity of the framework.