Tag
This paper introduces a semantic-timescale analysis pipeline to study how generic vs. specific content is distributed over time in human and AI-generated speech, revealing that autocorrelation-window measures capture temporal organization of semantics beyond static lexical distributions.