Moral discourse and stylistic critique at scale: Mapping themes and sentiment in Goodreads reviews of Uncle Tom's Cabin

Le résumé fourni par la source

Background Online reading communities such as Goodreads enable readers to publicly evaluate and interpret literary works at unprecedented scale, generating large bodies of reader-generated discourse that remain underexamined in literary scholarship. This study investigates contemporary reader reception of Harriet Beecher Stowe’s Uncle Tom’s Cabin (1852) to assess the extent to which modern readers engage with themes that have historically been emphasized by literary scholars. We hypothesize that contemporary readers continue to engage with the novel’s established themes, including slavery, Christian morality, and human suffering, and that positive reader sentiment is more strongly associated with moral and emotional themes than with other aspects of the text. Results To test these hypotheses, we analyze a corpus of 9,140 Goodreads reviews using computational text analysis. Topic modeling was conducted using Latent Dirichlet Allocation (LDA) and BERTopic to identify recurring thematic patterns, while sentiment analysis employed both the lexicon-based VADER model and the transformer-based RoBERTa model to evaluate reader attitudes toward the novel. The sentiment analyses indicate a predominantly positive reception overall, although RoBERTa identifies a larger proportion of negative and ambivalent evaluations than VADER. Topic modeling reveals recurring discussions of slavery and racism, Christian morality, emotional suffering, empathy, and the novel’s broader social and political significance, alongside discussions of narrative structure and reading experience. Conclusions The results provide support for the first hypothesis by demonstrating substantial continuity between contemporary reader discourse and themes documented in literary scholarship. They also offer limited or mixed support for the second hypothesis for the second hypothesis, as some of the most positively evaluated discussions center on the novel’s moral and emotional dimensions, although readers frequently combine ethical, historical, and aesthetic judgments in their assessments. More broadly, the study demonstrates how computational analysis of reader-generated discourse can complement traditional literary criticism by revealing large-scale patterns of interpretation and reception, highlighting both continuity and divergence between scholarly and public engagements with canonical literature.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.