Research And Implementation Of Narrative Visualization Video Automatic Generation Based On Data Fact

As data plays an increasingly important role in human life,it has become a new factor of production.Mining value from data rapidly and effectively can significantly improve individual and enterprise productivity.Narrative visualization can help people quickly discover interesting patterns and trends in data to understand the meaning and value of data better.Data videos,as one of the seven basic types of narrative visualization,are widely popular due to their vivid and high-density information advantages.However,existing data story video authoring tools require a combination of technologies and design capabilities.However,the recent data video authoring tools require the comprehensive use of multiple technologies and design capabilities,which cannot cover all steps of the visual data story creation process,making it almost impossible for ordinary users to create data videos independently.Besides,the platform’s threshold requirements for users to create data stories are too high,and its process of generating data story videos is not automated enough.Therefore,to address the above problems,this thesis proposes an automated pipeline process by absorbing automatic insight theory,and designs and develops an automatic visual data story video generation system for general users.The system integrates different stages of the visual data story video creation process.It automatically extracts meaningful data facts from the input spreadsheet,organizes them into easy-to-understand data stories by using sequence recommender,and synthesizes data story videos by automatically mapping dynamic charts.In order to provide an easy-to-use data story video generation software,the system is divided into four main functional modules:(1)In the automatic insight module,this thesis first designs ten types of data facts(such as value,trend,etc.).Then it propose s a calculation framework for data fact extraction combined with an importance calculation formula to ensure the quality of the data story.At the same time,a pattern extraction algorithm is proposed to extract different data facts,such as using Mann-Kendall’s trend test to extract trend-type data facts and Pearson analysis to extract correlation-type data facts.Finally,based on NLG research,a rulebased approach is adopted to design syntactic templates for different data facts to generate corresponding insight descriptions.(2)The sequence recommendation module implements the logical sorting and organization of the data facts sequences to support the creation of visual data stories.The sequence recommendation module identifies an optimal ranking for sequence recommendation through the sequence cost function of the Graphscape model,thus lowering the cognitive effort required by the target audience to understand and memorize the created visual data story.(3)Animation visualization module implements automatic animation generation of data-fact chart transitions based on four basic design principles for chart animation transitions.By secondary packaging of the Vizzu visualization framework,this module recognizes atomic transition operations from source visualization to target visualization and then gradually presents the chart animation process through phased animations,achieving the goal of using animation to guide the viewer’s attention and promote data perception.(4)The video compositing module applies Media Stream Recorder API and FFmpeg.wasm to realize video transcoding,audio uploading,and one-click sharing functions,making the data video complete and watchable.In summary,this thesis proposes a data video generation system that integrates data exploration and analysis,logical sorting of data facts,story script writing,and animation creation steps based on the shortcomings of existing data video creation tools.The system also provides automatic insight and sequence recommendation functions,making it easy for users to create interesting data stories with just one click from raw data.The development of this system effectively improves the work efficiency of data video creation and has a positive impact on improving the audience’s understanding and perception of data stories.