{Hiroya Takamura Makoto Miwa Anh-Khoa Duong Nguyen Mohammad Golam Sohrab}

Abstract

We present a neural exhaustive approach thataddresses named entity recognition (NER) andrelation recognition (RE), for the entity and relation recognition over the wet-lab protocolsshared task. We introduce BERT-based neuralexhaustive approach that enumerates all possible spans as potential entity mentions andclassifies them into entity types or no entitywith deep neural networks to address NER.To solve relation extraction task, based on theNER predictions or given gold mentions wecreate all possible trigger-argument pairs andclassify them into relation types or no relation.In NER task, we achieved 76.60% in terms ofF-score as third rank system among the participated systems. In relation extraction task, weachieved 80.46% in terms of F-score as the topsystem in the relation extraction or recognitiontask. Besides we compare our model based onthe wet lab protocols corpus (WLPC) with theWLPC baseline and dynamic graph-based information extraction (DyGIE) systems.