Together We Remember: Using Topic Modeling and Semantic Networks to Study Obituaries and Collective Memories of Contemporary United States

Obituaries have long served as an important source in recording memories and social values in the United States. However, prior research has focused primarily on celebrities or specific groups, often overlooking the narratives of ordinary citizens and neglecting broader discussions of their collective memories. This study addresses these gaps by applying large-scale data analysis and computational methods. We collected 39,449 obituaries from 2022-2023 and employed BERTopic topic modeling along with semantic network analysis to identify shared latent themes across individual obituaries. The findings suggest that, compared with celebrity obituaries, those of ordinary people are less scripted and more authentic in representing life details, yet still reveal identifiable latent "master narratives." Salient themes include family relationships, religion, military service, and alumni identity, alongside a noticeable shift toward portraying a middle-class ideal of life. Overall, this study contributes new insights to obituary research and introduces a novel framework for examining collective memory through large-scale commemorative texts.