Shakespeare Speeches with Rhetorical Figures Tagged (Miriam Joseph Catalogue)

This dataset provides a figure-by-figure rhetorical annotation of the Shakespeare canon, tagging 104,366 figure instances across 227 verified figure types in 38 plays — the 36 plays of the First Folio plus Pericles and The Two Noble Kinsmen — spanning 15,145 speeches. Each instance is labeled with its rhetorical figure (metaphor, antithesis, anaphora, epizeuxis, and so on) using the taxonomy codified in Sister Miriam Joseph's Shakespeare's Use of the Arts of Language (1947), the most comprehensive single inventory of Shakespeare's figures and one organized, uniquely, by the trivium itself. 95.9% of tagged instances correspond to figures Joseph defines; the remaining 4.1% are drawn from the broader classical tradition (Puttenham, Peacham, Lanham) under categories she does not treat. Annotation was produced by a multi-stage, two-pass large-language-model pipeline (Claude Sonnet, batch API). In the final stage, two model instances independently evaluated every speech by applying Joseph's classical definitions, and a third instance resolved disagreements by majority vote. The two primary instances agreed on 87.1% of individual tag verdicts without a tiebreaker. To anchor the pipeline against human judgment, a 200-speech gold-standard sample, drawn at random and hand-graded by the author applying the same definitions without access to the pipeline's verdicts, yields an F1 of 0.94 against the pipeline's final labels. To the author's knowledge this is the first canon-wide rhetorical annotation of Shakespeare checked against a human-graded gold standard: earlier annotation asserted its readings from a single pass, whereas this dataset adds the human-anchored verification of an automated pipeline.