Learning to Learn Recognising Biomedical Entities from Multiple Domains with Task Hardness

Anonymous

Abstract: Few-shot learning has been a big challenge for many classification tasks, where the final classifier is trained only with a few examples. This problem amplifies when we apply the few-shot setup to recognising named entity from different domains, i.e., few-shot domain adaption for NER. In this paper, we present a simple yet effective MAML-based NER model that can effectively leverage the task hardness information to improve the adaptability of the learnt model in the few-shot setting. Experimental results on biomedical datasets show that our model can achieve significant performance improvement over the recently published MetaNER model.