Interpreting Cultural Valuation Schemas with Computational Semantics

• In Event: Sociology of Culture Section Roundtables
  In Refereed Roundtable (90 minute): 9. Computational Sociology & Semantic Mapping

Abstract

This paper examines a corpus combining the Billboard Hot 100 Year-End pop lyrics dataset (1960-2020) with an additional series of popular songs extending back to the 1920s to evaluate changing American cultural reception schemas relating to social valuation over the twentieth century. We use contextual sequence embeddings of lyrics and apply various unsupervised methods for semantic change analysis including autoencoders, clustering, and semantic axis analysis to observe semantic shifts in the data, which we also briefly compare to embeddings of Google Books to test external validity. Across various measures of change, we find a general trend away from descriptions of external mediation of value, agency, and relationships and toward increasingly personally expressive or affective descriptions of these same constructs. We aim here to interpret whether this trend broadly mirrors widely observed shifts in valuation in the United States, which have been alternatively conceptualized as shifting from institutional to personally expressive modes of legitimation on the one hand and as extending institutional logics of valuation into interpersonal, affective, or extra-institutional domains on the other. Our findings more broadly aim to provide new context and methods for tracing historical and structural influences on macro cultural shifts in how valuation is constituted and interpreted.

Author

• William Rathje, University of California-Berkeley