TableIE: Capture the Interactions among Joint Information Extraction Explicitly via Double Tables
Anonymous
Unverified — Be the first to reproduce this paper.
ReproduceAbstract
Information Extraction (IE) mainly consists of three sub-tasks, Named Entity Recognition, Relation Extraction, and Event Extraction. Although the sub-tasks are highly correlated with each other, most previous works simply focus on part of them and ignore the interactions among different sub-tasks. In this paper, we propose TableIE, a double-tables based method to capture the interactions among joint information extraction (Joint IE, i.e., conduct all sub-tasks jointly) explicitly. Different from the complicated graph-based Joint IE methods, we maintain two tables for entity-relation extraction and event extraction respectively instead and implement the interactions among three IE sub-tasks explicitly. The experiment show that our TableIE model outperforms the previous state-of-the-art up to 2.5 in the ACE05 dataset.