Span-based Neural Model for Multilingual Flat and Nested Named Entity Recognition

Named entity recognition (NER) is an important task of finding entities with specific semantic types such as Protein, Cell, and RNA in text by addressing mostly with sequence labeling-based approach. In this paper, we present a bidirectional encoder representations from transformers (BERT)-based neural exhaustive approach that addresses flat and nested entities by reasoning over all the spans within a specified maximum span length. We evaluate the BERT-based span representation approach over the multilingual data sets of different domains including biomedical and general to address flat and nested entities. Experiment results show that span-aware-based approach is very effective to leverage NER.