Measuring Semantic Drift Across Generational Corpora: A Framework Using Pretrained Embeddings

Language evolves continuously, with words acquiring new meanings across generations. This paper introduces a systematic framework to measure semantic drift using pretrained embeddings applied to temporally segmented corpora. We present a proof-of-concept experi- ment comparing Wikipedia texts from two generational cohorts: Generation Z (1997–2012) and Generation Alpha (2013–present). Unlike resource-intensive diachronic embeddings, our approach leverages high-performance frozen models (OpenAI embeddings) to extract semantic representations efficiently. Analysis on ten conceptually charged words highlights measurable drift over time, underscoring the framework’s value for computational linguistics and digital humanities.