Editing Rhetoric (section)

== Automatic detection of rhetorical figures ==
As [[natural language processing]] has developed, so has interest in automatically detecting [[Rhetorical device|rhetorical figures]]. The major focus has been to detect specific figures, such as [[chiasmus]], [[Anaphora (rhetoric)|epanaphora]], and [[epistrophe|epiphora]]<ref>{{Cite journal |last=Dubremetz |first=Marie |date=May 2018 |title=Rhetorical Figure Detection: Chiasmus, Epanaphora, Epiphora |journal=[[Frontiers in Digital Humanities]] |volume=5 |doi=10.3389/fdigh.2018.00010 |doi-access=free}}</ref> using [[Statistical classification|classifiers]] trained with [[labeled data]]. A major shortcoming to achieving high accuracy with these systems is the shortage of labeled data for these tasks, but with recent advances in [[language modeling]], such as [[Few-shot learning (natural language processing)|few-shot learning]], it may be possible to detect more rhetorical figures with less data.<ref>{{cite arXiv |last=Madotto |first=Marie |date=August 2020 |title=Language Models as Few-Shot Learner for Task-Oriented Dialogue Systems |class=cs.CL |eprint=2008.06239}}</ref>