arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2211.07830cs.CL
收藏

提示语言模型以获取语言结构

Prompting Language Models for Linguistic Structure

Paul G. Allen School of Computer Science & Engineering(保罗·G·艾伦计算机科学与工程学院);University of Washington(华盛顿大学)

Terra Blevins, Hila Gonen, Luke Zettlemoyer

AI总结:

本文提出一种结构化提示方法,使自回归预训练语言模型能够进行零样本和少样本序列标注,实验证明其少样本性能优异,且表明PLMs的上下文学习与语言知识能力超越了训练数据记忆。

摘要:

尽管预训练语言模型(PLMs)可以通过提示执行广泛的语言任务,但这种能力在多大程度上源于可泛化的语言理解而非表层词汇模式,仍是一个悬而未决的问题。为测试这一点,我们提出了一种用于语言结构化预测任务的结构化提示方法,允许我们使用自回归PLMs进行零样本和少样本序列标注。我们在词性标注、命名实体识别和句子分块上评估了该方法,在所有情况下均展现出强大的少样本性能。我们还发现,尽管由于任务泄露到预训练语料库中,PLMs包含关于任务标签的显著先验知识,但结构化提示也能检索具有任意标签的语言结构。这些发现表明,PLMs的上下文学习能力和语言知识泛化超越了对其训练数据的记忆。

英文摘要:

Although pretrained language models (PLMs) can be prompted to perform a wide range of language tasks, it remains an open question how much this ability comes from generalizable linguistic understanding versus surface-level lexical patterns. To test this, we present a structured prompting approach for linguistic structured prediction tasks, allowing us to perform zero- and few-shot sequence tagging with autoregressive PLMs. We evaluate this approach on part-of-speech tagging, named entity recognition, and sentence chunking, demonstrating strong few-shot performance in all cases. We also find that while PLMs contain significant prior knowledge of task labels due to task leakage into the pretraining corpus, structured prompting can also retrieve linguistic structure with arbitrary labels. These findings indicate that the in-context learning ability and linguistic knowledge of PLMs generalizes beyond memorization of their training data.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑