Multi-Dimensional Visual Analysis Of Tang And Song Poetry Based On Temporal Attributes

Poetry is an important representative of Chinese traditional culture and classical literature,reflecting the author’s mood,social environment and other information.In order to deepen the understanding of poetry and explore the evolution of its creation,a multi-dimensional investigation with a macro perspective under the attributes of time is necessary,but traditional poetry research requires a lot of human and material resources due to technical limitations.Therefore,a set of multi-dimensional visual analysis methods based on temporal attributes is proposed for poetry of the Tang and Song dynasties.Firstly,existing authoritative data were collected to construct a collection of Tang and Song poetry to provide a reliable data source for the study.Due to the existence of dirty data in form and content,automatically check for null values,special symbols,etc.through traversal,and the anomalies in content are artificially corrected by consulting the data;by retrieving relevant information of the same subject from different data sources,three datasets containing temporal attributes of poetry,authors and historical events are integrated to avoid the dispersion of effective information.Secondly,in order to accurately predict poetic sentiment,BERT-CCPoem and Text CNN are integrated to build a multi-level sentiment model to better integrate global and local semantic features;themes are important for grasping poetic style,using LDA unsupervised clustering,a method is proposed to determine the number of themes by combining the mean value of theme similarity and the mean value of difference in document theme distribution,and adjust the number of themes in combination with the visual clustering effect;the traditional keyword extraction algorithm TF-IDF cannot meet the characteristics of poetry when extracting imagery,and the algorithm is improved by using category as the main body and probability instead of quantity statistics;considering the importance of imagery groups in poetry research,an imagery distance factor is added to the association rule mining algorithm Apriori confidence formula to reduce bias due to the uniform of imagery association degree.Then,in order to present the connection between multiple dimensions intuitively,the data information is mapped into visual elements through visual coding,and new visual charts such as bamboo forest chart,plum blossom chart,mountain and wild goose chart,temporal word cloud chart,and imagery chord chart are designed by using D3.js and Echarts to reduce the degree of fragmentation between dimensions;in order to present complex information flexibly,each chart is composed into a view through interactive design for visual analysis in a macro perspective;three research lines of poetry,theme and emotion,and imagery association are proposed according to the layout and interaction mode to give full play to the potential value of the view.Finally,in order to verify the feasibility and efficiency of the method,case studies and questionnaires are designed for each research line.Based on user feedback,existing methods were improved and future work was prospected.