Research And Implementation Of QA System Based On The Knowledge Graph Of Chinese Classic Poetry

With the rapid development of artificial intelligence technology,intelligent question answering(QA)systems have been widely used in various industries,such as intelligent customer service,intelligent AI assistants,etc.However,the existing intelligent QA system on the market for the field of Chinese ancient poetry application and research is very scarce.In recent years,with the goal of upholding cultural self-confidence,the concept of "culture is human and culture is moral" has been deeply rooted in people's hearts.At the same time,with the broadcast of popular variety shows such as "China Poetry conference","China's good poetry",people have become more and more enthusiastic about learning the cultural treasure of ancient Chinese poetry.In the traditional way,when people learn ancient Chinese poetry,they usually choose to use search engine to search for unknown problems,but the search results of search engine are mostly some web links or a paragraph of words.What's more,the search results of web pages are advertisements,which require users to spend a lot of time and energy to identify the correct information.Q & A system is more and more popular because it can understand users' intention quickly and provide users with high quality answers directly.The traditional question answering system is based on template matching or deep learning to generate question answers.In the face of increasing knowledge data,template matching based Q & A system needs to spend a lot of manpower and material resources to build a template database.The deep learning based Q & A system may produce answers that are not smooth or inconsistent with the intention of the question.While the Q & A system based on knowledge map provides high-quality data source for Q & A system by storing knowledge in the form of triple,and at the same time,it can provide accurate and concise results for users,it is more convenient for users to query and use,and attracts users to learn unknown knowledge.In view of the above situation,this paper uses the knowledge graph to study and realize the question answering system of ancient Chinese poetry.Through this system,the problems of ancient Chinese poetry can be accurately understood and mapped to the knowledge graph for solution.First of all,we need to crawl the knowledge of ancient Chinese poetry on the network through the crawler technology,and then store it in the graph database according to the concept layer design.Secondly,we use the framework of Rasa Q & a system to understand and answer the questions of ancient Chinese poetry put forward by users,and finally return the best matching answers to users through the human-computer interaction platform.Therefore,the main work of this paper is as follows:(1)This paper presents a question answering system model based on the knowledge map of ancient Chinese poetry.The model includes human-computer interaction module,question understanding module,question answering module and Chinese ancient poetry knowledge graph module.The human-machine interaction module is responsible for the interaction between users and the system.Users ask questions related to Chinese ancient poetry through the WeChat public platform,and the question answering system passes the questions answered to the users.The problem understanding module is mainly responsible for analyzing the natural language problems of ancient Chinese poetry put forward by users,and extracting the entities and intentions of the problems.The question answering module is mainly responsible for transforming questions into logical expressions,and connecting neo4 j Chinese ancient poetry knowledge map through cypher statements to query results.The knowledge map module of ancient Chinese poetry is mainly responsible for storing unstructured ancient Chinese poetry data into a triple knowledge expression form according to the concept layer design.(2)The construction of knowledge map of ancient Chinese poetry.Because there are different structures of ancient Chinese poetry knowledge on the network,and a large number of knowledge exists in unstructured data,so we need to take different technical methods to acquire and store knowledge.According to the characteristics of the ancient poetry website to be crawled,the knowledge of Chinese ancient poetry is acquired through the technology of web crawler.According to the concept layer design of the knowledge map of ancient Chinese poetry,it is stored in the form of triple,and the knowledge of different data sources needs to be integrated to realize the construction of the knowledge map of ancient Chinese poetry.(3)Understanding and answering the problems of ancient Chinese poetry.For the question of Chinese ancient poetry put forward by users,use rasa NLU module to understand the problem,and analyze the entity and intention of Chinese ancient poetry contained in the problem.Then use the rasa core according to rasa NLU understands the question intent and entity category,builds the corresponding graph database query statement to connect the knowledge graph for question solving,and uses mongodb to store the user dialogue history,which can better contact the context history information to improve the accuracy of question intent analysis when analyzing the user's question in the future,and finally returns the result to the user.(4)Implementation and test of Chinese ancient poetry Q & A system.According to the overall process and functional structure of the system,each module is coded and implemented.At the same time,a Q & A set of ancient Chinese poetry words is established as the test data to test the function of the Q & A system.Secondly,the Chinese ancient poetry Q & A system constructed is compared with the traditional search engine in terms of Chinese ancient poetry Q & A,and then tested with the commercial Chinese poetry Q & A system apple Siri in the field of Chinese ancient poetry Q & A.Finally,the Chinese ancient poetry quiz system is deployed to the Tencent cloud server,and the Chinese ancient poetry quiz service is provided through the official account of WeChat.In order to verify the validity of the q Q&A system of ancient Chinese poetry based on knowledge map,we run and test the Q&A system by collecting relevant questions of ancient Chinese poetry online.The experimental results show that for458 questions related to ancient Chinese poetry,the Q&A system constructed in this paper can correctly answer more than 92% of the questions.At the same time,compared with the traditional search engine and the commercial Chinese Q&A system,the test results show that the Chinese ancient poem Q&A system realized in this paper can make more effective and accurate answers in multiple rounds of dialogue and context connection,and the accuracy rate can reach more than 98%.It can effectively solve the lack of question answering system in the field of ancient Chinese poetry in the market,at the same time,it is convenient for users to query the relevant problems of ancient Chinese poetry anytime and anywhere,improve users' interest in learning and reduce learning costs.