Construction Of Subway Field Question Answering System Based On Knowledge Graph

With the development of AI and the implementation of the concept of Intelligent Transportation,the use of smart products to realize Intelligent Transportation has become a part of modern life.Especially for office workers and people traveling,there are generally subway or bus stations near companies and attractions.However,compared with buses,the most important thing about taking the subway is that there will be no traffic jams.Therefore,taking the subway for work commuting and travel has almost become the best choice.For people who travel,they are very unfamiliar with the subway lines in other places.They need to quickly understand the subway lines and surrounding information,as well as the elderly or children,who are inconvenient to type but need to query the subway information.Q&A system for subway filed based on knowledge graph are designed and implemented.Under the support of previous researches on Q&A system implementation and related technologies,an intelligent Q&A system based on subway filed knowledge graph is proposed,and its related theories and methods are researched and implemented.The main content can be divided into the following aspects:(1)Build a Q&A knowledge graph for subway filed.Firstly,this paper uses Web Scraper to collect subway data from the Internet,and does preprocessing,and then establishes and builds the ontology of the subway filed question and answer knowledge graph,and constructs a subway station information,attractions around each station,food,shopping information Subway Q&A knowledge graph and store it in Neo4 j graph database.(2)Knowledge reasoning is performed on the subway filed knowledge graph using a representation learning-based reasoning model.The knowledge reasoning model in this paper is established by Trans E,and realizes the reasoning of the potential relationship between the subway station and the surrounding attractions,food and shopping malls,and realizes the enrichment and expansion of the subway filed knowledge graph.(3)An intelligent Q&A algorithm based on subway filed knowledge graph is proposed.Named entity recognition algorithm and attribute linking algorithm are important foundations supporting the Q&A module.This paper proposes N-gram Mask Ro BERTa to realize the task of named entity recognition to complete the acquisition of question entities.Experiments show that N-gram Mask Ro BERTa has achieved good results in the subway filed named entity recognition task in the field of subway travel,with an F1 value of 92.46%,the highest score in the same group of experimental models.In this paper,the BERT model is proposed to implement the attribute linking task,and the bidirectional LSTM model is also selected for the same group of experiments,and the final scores of the two are compared.The results show that the BERT model can achieve better results in user intent analysis.(4)Design and implement an interactive subway filed Q&A system.It can support the basic information query of subway stations,and can give accurate answers to the questions raised by users in the subway field.