Multi-dimensional Visual Analysis Of Tang And Song Poetry Based On Space-time Attribute

The Tang and Song poems are the jewels in the crown of Chinese traditional literature,with extremely important historical,cultural and artistic values.Visual analysis technology,combined with deep learning and artificial intelligence can help people better understand and analyze the meaning of the poems.The existing research only presents the Tang and Song poems in static charts,which makes it difficult to obtain the correlation information between the multidimensional attributes of the poems.Lack of in-depth analysis of poetry,so a multidimensional visual analysis method based on temporal and spatial attributes of Tang and Song poetry is proposed.Firstly,this article combines the temporal and spatial attributes of Tang and Song poetry.Through a collaborative visual analysis method using multiple charts and techniques such as time filtering,map mapping,and quantitative statistical comparisons,it explores the macro trends of Tang and Song poetry evolution over time and geographical differences,taking into account multidimensional clues such as emotions,themes,content and quantity.The aim is to address the issue of loose relationships between charts,attributes,and data in the visual analysis process of Tang and Song poetry.Secondly,a multi-level emotion model BLATA based on BERT-Bi LSTM-Attention is proposed,which integrates the title of poetry and the author’s style.In the model,feature extraction is carried out on the text and title of poetry at the same time,so as to capture more abundant information of poetry.In the Attention layer,the characteristics of the writing style of poetry writers are coded into author vector,so as to obtain a more global impact of the writing style of poetry writers on the emotion of the whole poem,and solve the problem that the current emotion analysis model of poetry only pays attention to the text of the poem and ignores the title of the poem and the style of the author.The effectiveness of the model in introducing poetry titles and lyricist styles was evaluated by comparative experiments.Thirdly,on the basis of BLATA model,multi-head attention mechanism in CNN convolutional layer and Transformer model is introduced to build a BLMACTA model which can improve the extraction ability of local features and deep features.It can solve the problem of poor extraction ability of local features and deep features in the current poetry emotion classification model.The role of CNN convolution layer and multi-head attention mechanism in the model was evaluated by comparative experiments.Finally,a multi-dimensional visual analysis system of Tang and Song poetry based on temporal and spatial attributes is constructed.Through the linkage method within the chart,between the charts and between the multi-dimensional attributes,combined with a variety of interaction methods,it can not only help users obtain more vivid and effective poetry information in real time,but also deepen users’ understanding of the meaning of the poetry.Moreover,it can improve users’ perception ability of the evolution trend of Tang and Song poetry over time and the difference of regional distribution,providing new ideas for the future study of history and poetry.