Design And Implementation Of Library Human-Machine Dialogue System

In the routine work of a provincial library,readers often seek advice from library staff about library-related matters.On account of the large volume of such business,library staff and readers waste a lot of time.In the current development of computer technology and artificial intelligence,this library expects to implement a library human-machine dialogue system.The system which can introduce dialogue,one of the most common forms of communication between people,into the interaction between machines and humans,replace the work of library staff answering readers' questions,reducing the burden on library staff and providing readers with a better reading experience.According to the needs of the library,this paper designed and implemented a human-machine dialogue system,and finally deployed it to the library's servers to serve the library readers.According to the different function points,the system is divided into four modules,the most important part of which is the dialogue logic processing module for processing the dialogue logic,which is divided into the question answering system sub-module and chatbot sub-module.The question answering system sub-module is to meet the reader's consulting needs,in order to achieve this function defines a set of processes from pre-processing,keyword extension,classification of questions to similarity calculation and answer extraction.In the extension of keywords,word2 vec is used to obtain several words closest to the meaning of the target word as their synonym,and a variety of machine learning algorithms are compared in the problem classification and text expansion is used to improve the effect of the Naive Bayesian algorithm in short text classification.In similarity calculation and answer extraction,a variety of algorithm models are also integrated to improve the quality of problem matching and speed up query efficiency.The generative dialog model based on end-to-end and the dialog model based on template matching are fused in the implementation of the chatbot sub-module.The generative dialog module adopts the Seq2 seq model,designs several control experiments,analyzes the neuron types of recurrent neural network,attention mechanism and beam search algorithm on the generation effect of the recurrent neural network,and determines the generative dialogue model based on the analysis results.The dialogue model based on template matching is also used to improve the chatbot's performance on specific issues,improve the actual experience of the reader.In addition,it also actualized the proscenium module,management module and database module that are assorted with the dialogue logic processing module.Readers could use the web to access the system's services by the interaction between the modules.Although the system is based on the library business scene,but the implementation of the system possess outstanding universality,only need to make less changes to the system's metadata can be applied to other scenarios,the system is provided with both practical value and commercial value.