Research On Corpus Construction For Intelligent Response

With the maturity of natural language processing technology,there are more and more responsive robot products on the market.Among them,representative responsive robot products include Apple's Siri,Microsoft's Xiao Bing,Amazon's voice assistant Alexa,etc..The key content of R&D response robot is the construction of response corpus.At present,the construction methods of response corpus on the market mainly include retrieval construction method based on knowledge retrieval,situational construction method based on artificial template,and generation construction method based on deep learning.However,the response corpus constructed for a single knowledge-based retrieval construction,artificial template-based contextual construction,or deep learning-based generation method is far from achieving the desired effect,while the response content lacks professionalism in a certain field.Aiming at the current practice of constructing robot response corpus on the market is not practical,the expected effect is not achieved,and the response content lacks the common shortcomings of professionalism in a certain field.A hybrid intelligent response corpus construction method is proposed.The hybrid intelligent response corpus construction method proposed in this paper is based on the deep learning-based generative response corpus construction method combined with the AIML-based context-based response corpus construction method.That is to say,it has the accuracy of answering questions in the field of software engineering based on AIML-based context-based response corpus,and also has the professionalism(interpretation and memory ability for contextual content)in the daily response based on the deep learning-based generated response corpus.This paper is based on the deep learning learning response corpus construction method,which makes full use of the knowledge of the two-way LSTM coding model based on the Attention mechanism in deep learning to train the generated response model and generate daily conversations.At the same time,this paper focuses on the AIML contextual response corpus(based on The situational construction method of artificial template)can not support Chinese defects in a friendly way.By refactoring the AIML interpreter,AIML's contextual response corpus can better support Chinese,so that the software engineering question and answer content can be written into the AIML response library,and then The ability to answer questions in the field of software engineering.This paper writes this hybrid intelligent response corpus into two corresponding API interfaces(based on deep learning-based generated response corpus interface,AIML-based context-based response corpus interface),through the Android client and PC to the user show out.While ensuring the answering and daily response in the field of software engineering,the intelligent response corpus embeds the resource management corpus module to manage and manipulate the mobile phone resources,and then use voice to control and manage mobile applications,making the content of the intelligent response corpus richer..