Document-level Event Extraction - A Survey of Methods and Applications

Abstract

Event extraction aims to detect occurrences of specified types and extract corresponding event arguments from unstructured data input, which is an integral part of information extraction [1]. Many downstream tasks, such as text summarization, causal relationship identification, and event reasoning, make extensive use of event extraction. However, most of the existing research on event extraction remains at the sentence level. Document-level event extraction is still under exploration and lacks relevant review. Focusing on the task of document-level event extraction, this survey first divides the existing technologies into three categories and introduces the most representative models; Then, we list some commonly used datasets for document-level event extraction with their usage messages; Finally, we analyse the current research gaps of document-level event extraction and future research trends.

Streamlining event extraction with a simplified annotation framework

“…The event extraction task can generally be categorized into two groups: close domain and open domain (Xiang and Wang, 2019 ; Liu et al, 2021 , 2023 ). The close-domain extraction aims to extract a pre-defined structure based on supervised datasets.…”