Neural entity recognition with gazeteer based fusion
2021
Incorporating external knowledge into Named Entity Recognition (NER) systems has been widely studied in generic domain. In this paper, we focus on clinical domain where only limited data is accessible and interpretability is important. With recent advancement in technology and increased number of clinical trials has resulted in discovery of new drugs , procedures as well as medical conditions. These factors motivate towards building robust zeroshot NER systems. We propose an auxiliary gazetteer model and fuse it with NER system, which results in better robustness and interpretability across different clinical datasets. Our gazetteer based fusion model is data efficient, achieving +1.7 micro-F1 gains on i2b2 dataset using 20% training data, and brings +4.7 micro-F1 gains on novel entity mentions never presented during training. Moreover, our fusion model is able to quickly adapt to new mentions in gazetteers without re-training and the gains from name knowledge are transferable to related datasets.
Research areas