Abstract: How to associate data with stories and present data in the form of stories is the fundamental question in data story-telling research. Unlike previous approaches that employ a “hard connection” artificially separating data and stories, this paper adopts an integrated “soft connection” perspective, and proposes intermediate fusion elements (data facts and data events) to link data and stories. Based on these elements, a data storytelling model supporting quantification and automated operations is constructed. Firstly, leveraging Chatman??s narrative theory and utilizing linguistic features and natural language processing (NLP) tools, this paper concretizes and quantifies these fusion elements and designs an automatic recognition method. Subsequently, this recognition method is applied to extracting data facts and data events from interactive graphic works on Xinhuanet, posts on the Qingsongchou platform, and articles in Caixin’s “Data News” section. Analysis reveals a correlation between the positional distribution of these elements and the narrative structure, as well as identifying three narrative structures they constitute: vertical structure, horizontal structure, and interactive structure. Finally, a data storytelling model is constructed based on these data facts and data events. The results indicate that the proposed fusion element recognition method achieves Accuracy, Precision, and F1-score values over 0.8 on the three aforementioned datasets. Furthermore, the model constructed based on these fusion elements is present within data storytelling works and can be utilized to describe the generation process from data to story.

Key words: Chatman narrative theory, data storytelling model, narrative structure, data facts, data events

摘要： 如何关联数据与故事，将数据以故事形式呈现，是数据故事研究的基本问题。不同于以往的人为割裂数据与故事的“硬连接”，采用融合的“软连接”视角提出数据与故事关联的中间融合元素——数据事实和数据事件，并基于此构建了支持量化和自动化操作的数据故事模型。借助Chatman叙事理论，并依据语言学特征和自然语言处理工具，将融合元素具体化和量化，设计自动识别方法。应用识别方法提取新华网图文互动型作品、轻松筹平台发文和财新网“数字说”中的数据事实和数据事件，通过分析发现它们的位置分布与叙事结构存在关联，以及它们所构成的三种叙事结构（纵向结构、横向结构和交互式结构）。基于数据事实和数据事件构建数据故事模型。结果表明：所提融合元素识别方法在上述3个数据集上的Accuracy、Precision和F1-score得分均大于0.8；依据融合元素构建的模型存在于数据故事作品中，可用于描述从数据到故事的生成过程。

关键词: Chatman叙事理论, 数据故事模型, 叙事结构, 数据事实, 数据事件