Design And Implementation Of Cross-language News Event Graph Fusion

With the continuous improvement of artificial intelligence and the continuous increase of available data,knowledge graph and event graph have became a more successful application in the computer field.However,the vast majority of knowledge graphs and event graphs are constructed based on a single language,and one language cannot fully describe the knowledge of the whole world.In the era of big data,the global sharing of knowledge has became an international trend.Due to the problems of polysemy,polysemy and unregistered words among multiple languages,cross-language graph fusion has always been a research difficulty that is difficult to find the optimal solution.The purpose of cross-language event graphs fusion is find the event reference pointing to the same event in the real world in different language event graphs,which is helpful to complement and verify the cross-language event graphs.The thesis focuses on the construction of cross-language event graph fusion model,applies the graph convolution neural network algorithm to the cross-language event graph fusion model,and uses the event subject graph to represent the event center node and event related entity node information in the event graph,The fusion of cross-language event graphs is simulated as a graph matching problem between cross-language event subject graphs.The graph matching method is used to calculate the similarity between event subject graphs.Finally,the nodes of the two single language event graphs are aligned according to the matching state of the event center entity graph,and to complete the integration of cross-language event graphs.The main work of the thesis is divided into four parts:(1)Data collection: crawl structured and unstructured multilingual data through medical websites,news media,etc.as the data set of graphs construction.(2)News event extraction: extract and process the collected unstructured news data by building a news event extraction model,and convert it into structured data,and use it as the data set of graphs construction.(3)Construction of single language event graphs: build two single language medical news event graphs in Chinese and English to provide data support for graphs fusion.(4)Cross-language event graph fusion: firstly,build a cross-language event graph fusion model,and then apply the model to the constructed event graphs of two different languages,and output a cross-language event graph with high fusion degree.The thesis takes medical news as the research field,constructs a cross language event map fusion model,and tests the performance of the model.The thesis finally constructs a cross language medical news event map including Chinese and English,so that people can understand the news situation in the medical field from a multilingual perspective,and help people better analyze and predict the next news situation in the medical field.