Structuring Literature: Variants and Functions of Reflective Passages in Narrative Fiction

“所有幸福的家庭都是相似的，每个不幸的家庭各有各的不幸。”这是托尔斯泰的《安娜·卡列尼娜》的第一句话，也是虚构文学中被谈论最多的反思例子之一。尽管外行人也能将小说中的反思性段落与报道行为或描述人物的段落区分开来，但反思性段落在文学研究中尚未成为一个独立的研究课题。在此背景下，该项目将结合文学研究，语言学，计算模型和定量以及定性文本分析的见解，以(i)详细阐述和形式化反思性段落的综合概念，（ii）识别和分类叙事小说中的这种反思性段落，（iii）解释它们在大约350年的文学史上的发生模式。为此，该项目将以先前的工作为基础，自动识别其他文本类型中的反射，以及检测相关现象（如通用性和认识论状态），并将这些扩展并结合到一个语言学上知情的机器学习框架中，以寻找叙事小说中作者/叙述者归因于和角色归因于的反射段落。通过将我们的模型应用于跨越几个世纪的德国文学的大量语料库，我们将能够确定反思特别普遍的时期（“繁荣时期”）。我们还将更密切地关注文学史上两种突出的反思性叙事变体：散文式叙事和百科全书式叙事。我们将比较叙事性小说中自动识别的散文式和百科全书式的段落与非虚构的文章和百科全书式的文本，并研究在风格、主题和句子的认知地位方面的异同，以确定这些段落与文学作品及其社会文化背景有关的内部和外部功能。我们对整个优先项目“计算文学研究”的贡献包括：(i)通过进行定性和定量的语料库研究来实现文学理论的基本概念；（ii）丰富构建文学文本的方法选择；（iii）开发新的计算模型，用于检测反思和相关现象，并展示这些算法方法对文学史编纂的有用性，以及将其结果整合到定性研究过程中；（iv）创建并发布手动注释反思的语料库。

"All happy families are alike, each unhappy family is unhappy in its own way," is the first sentence of Tolstoy's "Anna Karenina" and one of the most talked about examples of reflection in fictional literature. Although even lay people distinguish such reflective passages in novels from passages that report actions or describe characters, reflective passages have not yet been established as a research topic in their own right in literary studies. Against this backdrop the project will combine insights from literary studies, linguistics, computational modelling and quantitative as well as qualitative text analysis to (i) elaborate and formalise a comprehensive concept of reflective passage, (ii) identify and classify such reflective passages in narrative fiction and (iii) explain their patterns of occurrence in roughly 350 years of literary history. To this end, the project will build on previous work for identifying reflection automatically in other text types as well as work on detecting related phenomena such as genericity and epistemic status and extend and combine these in a linguistically informed machine learning framework for finding author/narrator-attributed and character-attributed reflective passages in narrative fiction. By applying our models to a large corpus of German literature spanning several centuries, we will be able to identify periods in which reflection was particularly prevalent (“boom periods”). We will also look more closely at two prominent variants of reflective storytelling in literary history: essayistic and encyclopedic narration. We will compare automatically identified essayistic and encyclopedic passages in narrative fiction to non-fictional essays and encyclopedic texts and investigate similarities and differences with respect to style, topic and the epistemic status of clauses in order to identify internal as well as external functions of these passages in relation to the literary works and their socio-cultural context. Our contribution to the overall Priority Program “Computational Literary Studies” consists in (i) operationalizing a basic concept from literary theory by performing a qualitative and quantitative corpus study, (ii) enriching the methodic options to structure literary texts, (iii) developing novel computational models for detecting reflection and related phenomena and demonstrating the usefulness of these algorithmic methods for literary historiography as well as the integration of their results into the qualitative research process, and (iv) creating and releasing a corpus manually annotated for reflection.