Multidimensional Visual Analysis Of The Theme And Character Relationships In Chapter-based Novels

As a literary form,chapter-based novels contain intricate storylines and character relationships.To gain a deeper understanding of their essence and value,exploration from multiple dimensions is necessary.However,traditional research methods face technical limitations and high costs.Visual analysis can aid researchers in delving into the structure and content of these novels.Existing visual analysis methods,however,fail to comprehensively explore their unstructured data.Therefore,the research will focus on the following aspects.Firstly,the Python urllib module will be utilized to crawl textual data of chapter-based novels.The text will undergo data cleansing using Jieba word segmentation and a stop-word dictionary.The CRF method will be employed to extract identifiers such as character names,time,and locations from the novels’ texts.These identifiers will be combined with geospatial data from the Chinese Geographical Information System database(CHGIS),character relationships from CN-DBPedia,and story background data.After filtering,a multidimensional dataset of chapter-based novel information will be obtained,consisting of approximately 5000 records containing information on characters,locations,and events.Secondly,to address the issue of Text Rank algorithm overlooking implicit variables in themes,a hybrid topic extraction algorithm combining LDA and Text Rank will be proposed.The topic word weights obtained from LDA will serve as the topic influence factor,initializing the weight values of vertices in the Text Rank algorithm.This approach aims to enhance the accuracy of topic extraction and will be evaluated through parameter comparison experiments involving LDA,Text Rank,and the proposed hybrid algorithm to validate its effectiveness.Next,a character relationship analysis method based on an improved Node2 vec algorithm and hierarchical clustering will be proposed to tackle the complex character relationships in chapter-based novels.Considering the influence of character importance on relationships,a weight factor will be introduced into the Node2 vec algorithm to improve the accuracy of character relationship differentiation.To address the issue of unclear presentation of complex relationship displays in Gephi,a combination of K-means and hierarchical clustering will be used to optimize the clustering effect of the relationship graph.Finally,a multidimensional visual analysis method will be designed with a focus on analyzing the chapter-based novels of Jin Yong.The method will involve analyzing the source text from various perspectives,including theme plots,story environments,and character relationships in chapter-based novels.Visual analysis methods such as character clustering views,character statistics views,thematic temporal views,narrative views,and knowledge graph views will be designed to present the dataset and experimental results in a multidimensional manner.Interactive techniques will be employed to achieve a collaborative visual analysis effect across multiple views.To validate the feasibility and efficiency of the proposed methods,case studies and user feedback will be conducted.Based on user feedback,improvements will be made to the existing methods,and future work will be discussed.