Visualanalysis Of Multidimensional Evolution Of Poetic Themes And Poetic Portraits

Poetry,as the main carrier of excellent culture,has distinct thematic and contemporary characteristics.The existing research on poetry usually revolves around the Tang and Song dynasties,focusing on themes,artistic conception,and the poet’s life,neglecting the changes between dynasties and the evolution of thematic content.Therefore,starting from the poetry of the Pre Qin to Qing dynasties,this study aims to explore the thematic characteristics and evolution of each dynasty based on the division of dynasties,explore the era characteristics and writing rules contained in poetry,and draw poetic portraits based on themes and poems,providing a means for literary researchers and poetry enthusiasts to assist in the study of poetry.Firstly,Beautiful Soup method is used to obtain poetry data,and Pandas is used to preprocess the data by removing duplicates and missing values.Then,a full dynasty poetry text dataset containing multidimensional attributes is constructed;CRF method is used to analyze the creation background of poetry,extract the writing time of poetry,and supplement the time attributes of poetry with the Tang and Song literary chronicle map.After filtering,a poetry time dataset with creation time is obtained;Py Query method is used to obtain historical events from the all history website.Secondly,Jiayan and stop words dictionaries are used to segment and remove stop words from poetry of different dynasties,and LDA topic model is used to extract poetry themes from poetry of different dynasties after word segmentation;TF-IDF method is used to extract the probability distribution of poetry themes,and the setting method of theme threshold based on the probability distribution of themes is proposed to calculate the multi-theme distribution of poems.Then,since CLIP model is only suitable for modern corpus,aiming at the research of ancient poetry,a training corpus of 50 000 pieces of ancient poetry translation is constructed by taking the translation of ancient poetry website and ancient books as the main source.Encoder-Decoder model is used to translate ancient poetry and obtain the translation of ancient poetry.In order to vividly express the theme and artistic conception of poetry,VQGAN model combined with Codebook and Transformer is used,and the translation of ancient poetry is taken as the input of CLIP method to guide VQGAN model to match the poetry and image pairs,and generate poetic portraits consistent with the theme and artistic conception.Finally,according to the requirements of visual analysis,a series of new charts such as theme ring chart and theme landscape chart are designed and implemented to explore the theme and era characteristics of poetry,and explore the evolution of themes and the differences in poetry content from multiple dimensions such as historical events,poetry creation time and era background.The theme and artistic conception of poetry can be conveyed more intuitively and accurately with the help of poetic portraits.Using a questionnaire survey form for case analysis and visual analysis evaluation.The results show that the multidimensional evolution of poetry themes and visual analysis of poetic portraits have good effectiveness and practicality,which can help literary researchers and poetry lovers to effectively analyze the themes and images of poetry.