Ellipsis is a linguistic phenomenon that occurs in various languages, including Chinese. Although humans can generally understand text with omissions correctly, it can have an impact on machine understanding in terms of syntax and semantics. Therefore, the automatic recovery of omitted elements is of significant importance for automated text analysis and comprehension. This task proposes a computationally feasible omission recovery task that aims to restore omitted constituents that occupy valid positions in the syntactic structure of a sentence while playing a semantic role. The task is divided into two subtasks: ellipsis position detection and ellipsis content generation. Baseline methods that have achieved good results in both subtasks are described. Additionally, to advance research on large language models, this study also attempts to utilize the approach of in context learning using ChatGPT to perform this task and conducts relevant analysis.
关键词
省略探测/省略恢复
Key words
省略探测/省略恢复
引用本文复制引用
会议名称
Chinese national conference on computational linguistics
会议地点
Harbin(CN)
会议母体文献
22nd Chinese national conference on computational linguistics (CCL 2023): evaluations
NETL