基于LDA主题建模的文学语料主题演化分析——以简·奥斯汀六部小说为例

Topic Evolution in Literary Corpus Based on LDA Topic Modeling—A Case Study of Six Jane Austen Novels

摘要: 为厘清简·奥斯汀创作主题的阶段性变化，并建立文本词汇特征与时代社会环境的关联，本文选取其6部长篇小说构建小型英文语料库，采用LDA主题建模开展量化文本分析，并结合19世纪初英国文学发展脉络完成综合解读。量化结果显示：奥斯汀早期创作文本高频词汇集中于情感、社交婚约相关表述，文本整体侧重婚恋关系中的情感与礼仪探讨；成熟期作品高频出现财产、生存处境类词汇，文本更侧重个体婚恋选择背后的现实约束。分作品独立建模能够捕捉同一创作阶段下不同文本的微观词汇差异。统计层面呈现的主题阶段性分化，主要源于当时英国财产继承、婚姻财产相关法律对女性生存选择的硬性约束，社会财富流动变化仅作为长期时代背景存在。本研究构建了“量化数据 + 文学解读 + 社会语境”的融合框架，搭建“量化词汇统计 + 文本细读 + 时代制度语境”的综合分析路径，验证了LDA模型在小规模文学语料库中作为统计发现工具的辅助性应用价值，为计算语言学与文学研究的交叉提供了可复用的分析路径。

Abstract: To clarify the phased shifts in Jane Austen’s creative themes and establish correlations between textual lexical features and the social context of her era, this paper constructs a small English corpus consisting of her six full-length novels and conducts quantitative textual analysis via LDA topic modeling, followed by an integrated interpretation against the literary context of early nineteenth-century Britain. Quantitative findings indicate that high-frequency lexical items in Austen’s early works revolve around emotions, social etiquette and marital engagements, with narratives centering on emotional bonds and ritual norms within romantic relationships. Her mature writings, by contrast, feature frequent terms pertaining to property and living circumstances, placing greater emphasis on practical restrictions underlying individuals’ marital decisions. Independent modeling of individual novels further captures micro-level lexical disparities within the same creative phase. Statistically evident thematic segmentation primarily stems from rigid legal constraints imposed by British laws governing property inheritance and spousal assets on women’s life choices, while shifts in social wealth circulation merely function as a long-term historical backdrop. This study establishes an integrated analytical framework combining quantitative data, literary interpretation and social context, alongside a comprehensive research path integrating lexical statistics, close reading and institutional context of the age. It verifies the auxiliary value of the LDA model as a statistical auxiliary tool for small-scale literary corpora, and delivers a replicable analytical workflow for interdisciplinary research bridging computational linguistics and literary studies.