Cross-media Science And Technology Big Data Knowledge Graph Construction And Dynamic Precision Portrait

With the exponential growth of the scale of scientific and technological resources,the existing scientific and technological resources are facing problems such as numerous indicators,subdivision of categories,difficulty in complete coverage,and precise refinement.Combining artificial intelligence with the field of science and technology,with the help of knowledge graphs to achieve high-precision portraits of science and technology resources,has broad application prospects and practical significance.Based on the characteristics of science and technology big data,this thesis conducts the collection,processing and storage of cross-media science and technology big data,the construction of cross-media science and technology big data knowledge graph,the dynamic and accurate portrait research of cross-media science and technology big data,and finally realizes the cross-media science and technology big data Knowledge graph construction and dynamic accurate portrait system.The specific contributions of this thesis are as follows:(1)In terms of the collection,processing and storage of cross-media science and technology big data,in view of the characteristics of cross-media science and technology big data,a method for obtaining unstructured and structured data is proposed.Use major network information disclosure platforms and distributed crawler tools to obtain data,and at the same time build a corresponding cross-media science and technology big data lexicon for data noise reduction and normalization,and finally use MySQL relational database for effective data storage.This thesis has obtained a total of 263,725 original policy data from 23 provinces including "Beijing","Shanghai" and "Guangzhou".(2)In terms of the construction of the knowledge graph of cross-media science and technology big data,based on the current research status in the field of knowledge graphs and the characteristics of the text data of cross-media science and technology big data,this thesis will construct the knowledge graph from the original unstructured text data.Among the key technologies,the entity recognition algorithm based on BERT-BLSTM-CRF and the entity relationship extraction algorithm based on BGRU-BATTENTION are proposed.Specifically,this thesis uses the two key technologies of named entity recognition and entity relationship extraction to construct a cross-media science and technology big data knowledge graph.At the same time,this thesis completes data storage and visual analysis in the graph database Neo4j.Among them,in terms of named entity recognition,the entity recognition algorithm based on BERT-BLSTM-CRF proposed in this thesis has an accuracy rate of 3.69%higher than that of current competing algorithms on cross-media science and technology big data data sets.At the same time,the BGRU-BATTENTION-based entity relationship extraction algorithm proposed in this thesis has an accuracy rate of 2.65%higher than that of similar competition methods in cross-media science and technology big data..(3)In terms of dynamic and accurate portraits of cross-media science and technology big data,due to the changes in source data,the improvement of extraction requirements,and the refinement and enrichment of information,this thesis uses automatic crawling of new data based on time series to complete the data The superposition of information and the expansion of graphs are based on traditional machine learning methods to resolve entity ambiguities caused by data increments to ensure the accuracy of big data portraits.Finally,based on the time dimension,we constructed different time series of cross-media science and technology big data accurate portraits,and completed the construction of the time dimension of cross-media science and technology big data dynamic and accurate portraits.(4)Designed and implemented the construction of the "Knowledge Graph Construction and Dynamic Accurate Portrait of Cross-Media Science And Technology Big Data" system.Through the collection,processing and storage module of cross-media science and technology big data,the building module of cross-media science and technology big data knowledge graph and the dynamic and accurate portrait module of cross-media science and technology big data,the data for cross-media science and technology big data is realized.Collection and processing,knowledge map construction,dynamic accurate portraits and other functions.At the same time,a user-friendly algorithm interface and a simple and clear visual interactive interface are designed,and the results of the proposed algorithm model are fully displayed.