The article introduces a new task in the field of natural language processing – rhetorical parallelism detection. The authors provide a formal definition of the task, establish a family of metrics to evaluate performance on it, and create baseline systems and novel sequence labeling schemes to capture it. They also provide a Latin and a Chinese dataset for the task. The study is a first in the computational modeling of parallelism, a common stylistic tool in rhetoric that has seldom been investigated in natural language processing. The task aims to better understand the structure, meaning, and intent of human communication.
Publication date: 30 Nov 2023
Project Page: https://arxiv.org/abs/2312.00100v1
Paper: https://arxiv.org/pdf/2312.00100
