基于监督、零样本和少样本的立场检测应用（英文版）

英文标题：Stance Detection With Supervised, Zero-Shot, and Few-Shot Applications中文摘要：本文通过确定立场检测任务的定义，提供了一个通用框架，并提供了三种不同的方法来执行立场检测：监督式分类、NLI 分类器的零样本分类和上下文学习。此外，本文论述了零样本和少样本语言分类器的应用和局限性与监督分类器的不同，并通过复制 Block Jr 等人 (2022) 的方法演示了零样本立场检测的应用。英文摘要：Stance detection is the identification of an author's beliefs about a subjectfrom a document. Researchers widely rely on sentiment analysis to accomplishthis. However, recent research has show that sentiment analysis is only looselycorrelated with stance, if at all