Research On Multi-Dimensional Knowledge Organization And Visualization Of Records Of The Grand Historian

The vast historical books carry the long history of China,are valuable materials for understanding the ancient Chinese society.In the aspect of genre feature,the Chronicle books take time as the mainline to present history.It is easy to reflect the connection between various historical events that happened at the same time,but difficult to reflect the people’s deeds related to significant historical events.In the western Han dynasty,Sima Qian has created a new genre,which present the history in a series of biographies.Since then,including The Records of the Grand Historian(Shi Ji)and The History of Han Dynasty(Han Shu),the twenty-four histories all adopt this genre.But it is difficult to show the connection between different events simultaneously in a biographical history book.In addition to narrative structure history books,historical knowledge in history books can also be presented from multiple dimensions,such as characters,time and place,etc.,which can not only make up for genre defects to some extent,but also facilitate readers to select corresponding dimensions to obtain information according to their own needs.However,it takes a lot of time and effort to deal with the existing history books manually.Based on the development of digital humanity,this paper takes Shi Ji,the first historical book which presented the history in a series of biographies in Chinese history,as the experimental data.Based the theory and method of Information Science and Computer Science,this article achieves the goal of re-organizing and visualize the knowledge in Shi Ji by following four steps:(1)Knowledge modeling of Shi Ji and basic corpus constructionThe multi-dimensional knowledge modeling of history books and the construction of basic corpus are the basis for the development of history books centering on characters,time,place and other dimensions.In this chapter,based on SPO-X,a knowledge representation model,SPO-TS was defined,the multi-dimensional knowledge modeling of Shi Ji is completed.According to the knowledge modeling results,the full text and entity level resources related to Shi Ji are obtained and preprocessed,which has provided a guarantee for the follow-up research.(2)SPO knowledge extraction system for history booksThe knowledge extraction of historical records is the key technique for knowledge in history books to present according to characters and places.Based on the statistical analysis,this research improve BERT model by adding word segmentation and part-of-speech features.And adopt that the method of pipeline method to construct knowledge extraction system.Firstly,identifying multiple types of knowledge(properties)contained in sentences.Secondly,extracting Subject(Object)pairs of entities involved in knowledge.The Precision of the system is 70.07%,the Recall is 65.78%,and the F-score is 67.86%.(3)Generating Time Evolution Sequences of Historical EventsAutomatically generating time-evolution series is the key technique for knowledge in history books to present according to time.For ancient Chinese texts,the time expression extracted from the text cannot be used directly.There are 4 situations:Firstly,element of time incomplete;Secondly,lots of sentences inherit time from the previous section.Thirdly,the ambiguity of time expression.Lastly,different time basis points.In this chapter,we explored the time evolution sequences of historical events and break the task down into time expression recognition,time expression normalization,time linking and event time alignment,.(4)Visualization of multi-dimensional knowledge in Shi JiOn the basis of the above research,a multi-dimensional knowledge visualization platform of is constructed to present the historical knowledge of Shi Ji from the three dimensions,namely time,character and place.