Design And Implementation Of Subject Knowledge Base Supporting Semantic Reasoning

In 2018,according to statistics from the National Science Foundation(NSF),the total amount of Chinese scientific publications surpassed the United States for the first time,as the most representative indicator of innovation capability,reflecting the economic development and society of a country.Rapidly growing data and accurately identifying its corresponding subject areas,whether it is knowledge category or high-efficiency search,have great strategic significance.The knowledge base system not only saves time in retrieving relevant subject knowledge in terms of efficiency,but also has commercial potential that cannot be underestimated in the future.The knowledge base provides full-text search service for papers and supports automatic classification of disciplines.According to the construction process of knowledge base,it can be divided into six core modules,namely data support module,subject classification module,knowledge information retrieval module,statistical analysis module,semantic reasoning module,log monitoring module.The main contents are as follows:(1)Data collection,using the billion-level data set combined with crawler data as the knowledge base data foundation,based on this,effective data cleaning,integrated into the library search engine,providing full-text search function.(2)Subject classification,through the pre-training word vector,combined with the convolutional neural network to train the model,and compare FastText to classify the data text.(3)Knowledge information retrieval,to achieve a variety of search methods,including simple search,advanced search,domain search,for the search results can be sorted according to the sorting method,export and other functions.(4)Statistical analysis,providing a rich visual display of the search content,including global statistical analysis and retrieval statistical analysis,using different analysis methods according to the search content.(5)Semantic reasoning,based on data support,to achieve named entity recognition and relationship extraction.(6)Log monitoring,monitoring node and index level of the cluster,configuring early warning rules,quick log viewing,and cluster bottleneck monitoring.The system adopts the B/S architecture as the core built Web service platform.In the front-end display layer,jQuery and Echarts are used as the core framework,the logical layer adopts Django as the business control center,and for the data layer,Elasticsearch is adopted as the storage and retrieval platform,and Data support is provided in conjunction with Redis non-relational database,Kibana as a monitoring tool,model training using Keras deep learning framework,word vector training using Gensim,contrast model using FastText framework for comparison,entity recognition and relationship extraction using Standord CoreNLP,project development stage,The Huawei DevCloud platform is used as a version control tool.During the test phase,Esrally is used to perform the search function test,and Selenium is used for automated testing to detect compatibility issues.Since the project was established,the company has developed and launched the first version.All the functions of the knowledge base system have been completed.During the trial operation phase of the system,the developer will be able to expand and stabilize the system according to the user feedback.Effectively improve and perfect the aspects of sex,and gradually build a platform that can better meet the needs of users,so that the platform can be promoted and used in a wider range.