Research And Implementation Of Film Knowledge Question Answering System Based On Knowledge Graph

With the development of the Internet,the content of network data presents an explosive growth trend.Due to the characteristics of large amount of data,multiple types of information and inconsistent structure of Internet content,people are challenged to acquire information and knowledge effectively.Knowledge Graph,with its powerful semantic processing ability and open organizational ability,has laid the foundation for knowledge organization and intelligent application in the Internet era.As a novel knowledge organization and retrieval technology in the era of big data in the past two years,knowledge atlas has gradually shown its advantages in knowledge organization and display,and has attracted the attention of many industries.Knowledge graph is used to represent the relationships between entities in the real world.With the development and application of artificial intelligence technology,knowledge graph has become one of the key technologies in intelligent search,automatic question and answer,personalized product recommendation and other fields.At present,China's film industry is developing rapidly,and the demand for watching movies continues to expand.But the way for users to effectively get movie information is still search engine and professional movie website.It is not friendly for users who want to quickly know the film-related information or search for the film individually according to the conditions.Therefore,this paper constructs a knowledge graph of the film field with complete information,and realizes an automatic question-and-answer system of film knowledge based on template matching.The specific research work of this paper is as follows:(1)Using the browser development tool to grab the package test,we can find the JSON web page link of Douban Movie Data,and use the combination of requests,bs4 and regular expression to grab Douban Movie Data and store it in Mysql database.IP pool and random user-agent are used to circumvent the anti-crawler restriction of websites and build multi-threaded web crawler tools to improve the efficiency of crawlers.According to the captured Douban Movie data,the entities,relationships and attributes of the Movie Knowledge Map are designed,and then the captured data are processed and imported into Neo4 j,the Map Database,and the Movie Knowledge Map with more comprehensive information is constructed.(2)Designing movie knowledge query template and building user question training set,using TF-IDF algorithm to extract text features and train naive Bayesian question classification model;crawling movie news data,tagging movie name entity corpus data set,adding pre-trained Word2 Vec word vector language model to BiLSTM-CRF named entity recognition model,The recognition effect of named entity recognition model is improved.(3)In the final answer query task,a fuzzy query based on Bidirectional Maximum Matching is proposed,aiming at the problem that user input questions may have errors in character recognition and named entity recognition,and the output results are sorted by calculating cosine similarity,which improves the practicability of the system.Finally,a web system display platform based on flask is implemented.