Research And Implementation On A Chatbot System

In recent years,with the development of speech recognition and natural language processing technology,chatbots have been regarded as the future man-machine interaction because its friendly experience.However,chatbots based on knowledge extraction models or generation models are far from what people desire.With the outbreak of Web services/APIs,we propose a chatbot based on service matching.The system responds to the user's input by calling web service,with the Chinese natural language processing technology for service discovery and matching.The system can solve the problem that the current chabots have.We propose a service matching algorithm for chatbots.The keywords and named entities are extracted from the text entered by user through Chinese word segmentation,POS tagging,and named entity recognition.Then,the keywords are used to retrieve API,and the extracted keywords and entities are matched to the API's parameters.In this paper,CRF-BiLSTM model is used for Chinese words segmentation.The model is trained based on the SIGHAN Bakeoff 2005 data set,and the experimental results are analyzed.The results show that compared with the dictionary-based mechanical word segmentation method this method increase nearly 10%on accuracy,recall.rate and F value.This model have reached the best level of Bakeoff2005 Competition Open Group.We propose RESTbot architecture style for chatbots.Implemented based on REST architecture style,RESTbot develops a further specification for the using of bots.Developers do not need to code to transform the data after bots calling RESTbot style Web Services.Web services can be directly visualized based on the type of message,which reduces the repetitive and meaningless data transformation work.Finally,We implement a chatbot using micro-service architecture.We introduce the micro-service architecture,API gateway,service management and Web service in our system.