Research And Implementation Of Question Answering System For Intelligent Tour Guide Robot

China’s tourism resources are rich and diverse.With the growth of tourist scale,more and more scenic spots use guided robots to improve service efficiency.Question answering system is the key to human-computer interaction of robots,and knowledge retrieval model is the main research content of question answering system.The existing knowledge retrieval models are mainly divided into text retrieval model(TR)and knowledge graph embedding model(KGE).Relatively speaking,the retrieval ability of KGE is superior to TR,which is the mainstream model.However,it also faces the following problems:First,the storage and retrieval ability of knowledge graph for massive factoid information is superior to the text retrieval model,but KGE lacks the ability to infer non-factoid information;second,the KGE model ignores the semantic information contained in the process of processing factoid information.Therefore,this thesis focuses on the development of a Q&A system for intelligent tour guide robot,which can interact with tourists about the scenic area content and basic tourism knowledge.It mainly includes:1.This thesis design and implement a knowledge graph embedding that integrates semantic information(SIKGE),and achieve the balance between triplet structure information mining and semantic information mining by introducing a pretraining model;2.After the implementation of SIKGE,this thesis design and implement a knowledge retrieval model based on non-factoid information enhancement(NFE-KRM).It realizes the semantic information extraction model SFEM to extract the semantic space expression of non-factoid information,and the Unified Semantic Embedding Space(USES)is realized by constraining L2 distance.So that the knowledge retrieval model has the ability to answer both factual and non-factoid questions.3.On the basis of the implementation of NFE-KRM,this thesis design and implements a robot question answering system,which supports voice users’interface(VUI)and graphical users’ interface(GUI)in combination with the question-answer documents provided by the scenic spot and the open source knowledge graph.This thesis has used a large number of experiments to prove that SIKGE gets a better performance on Mean Rank and Hits@10.NFE-KRM’s F1 score and accuracy on the mixed dataset are both competitive.The intelligent tour guide robot with the question and answer system implemented in this thesis has been tested in Guizhou Shenquan Valley scenic spot and other places.It can save operating costs and improve service efficiency,also can be converted in scenic spots,shopping malls,games and other scenes at low cost.What’s more,in the light of the robot had generated certain economic and social benefits,2022 China International Fair for Trade in Services(2022CIFTIS)invited it to participate in the exhibition.