A Bidirectional Iterative Algorithm for Nested Named Entity Recognition

Abstract

Nested named entity recognition (NER) is a special case of structured prediction in which annotated sequences can be contained inside each other. It is a challenging and significant problem in natural language processing. In this paper, we propose a novel framework for nested named entity recognition tasks. Our approach is based on a deep learning model which can be called in an iterative way, expanding the set of predicted entity mentions with each subsequent iteration. The proposed framework combines two such models trained to identify named entities in different directions: from general to specific (outside-in), and from specific to general (inside-out). The predictions of both models are then aggregated by a selection policy. We propose and evaluate several selection policies which can be used with our algorithm. Our method does not impose any restrictions on the length of entity mentions, number of entity classes, depth, or structure of the predicted output. The framework has been validated experimentally on four well-known nested named entity recognition datasets: GENIA, NNE, PolEval, and GermEval. The datasets differ in terms of domain (biomedical, news, mixed), language (English, Polish, German), and the structure of nesting (simple, complex). Through extensive tests, we prove that the approach we have proposed outperforms existing methods for nested named entity recognition. INDEX TERMS information extraction, natural language processing, nested named entity recognition,

The Chilean Waiting List Corpus: a new resource for clinical Named Entity Recognition in Spanish

“…In terms of developing NER models trained on this corpus, future work includes improving the recall for disease and medication, due to the importance of identifying these entities, and dealing with nested NER, for which there is a variety of approaches as summarized in (Dadas and Protasiewicz, 2020). Besides, our annotated corpus has hierarchical entities (for example, test result and sign/symptom are part of the entity finding).…”

LIPN at WojoodNER shared task: A Span-Based Approach for Flat and Nested Arabic Named Entity Recognition

“…Deep learning techniques, particularly recurrent neural networks (RNNs) and recently, transformer-based architectures like BERT (Devlin et al, 2019) and GPT (Radford et al, 2019), revolutionized NER. These models leverage contextual embeddings to capture intricate relationships and dependencies, achieving state-of-the-art results in various languages and domains for both flat (Xia et al, 2019;Zheng et al, 2019;Arkhipov et al, 2019;Lothritz et al, 2020;Yu et al, 2020;Yang et al, 2021) and nested (Sohrab and Miwa, 2018;Katiyar and Cardie, 2018;Dadas and Protasiewicz, 2020;Wang et al, 2020) entities.…”

S-NER: A Concise and Efficient Span-Based Model for Named Entity Recognition

“…We conduct experiments on the dev sets of the three benchmark datasets to investigate the effects of the negative sampling strategy. We set the negative span count δ to different values, i.e., 0, 2,4,6,8,10,20,30,40,50,100,150,200, 250 and 300. We report the experimental results in Figure 3, from which we observe that a sufficient number of negative spans is essential to ensure that the model performs well.…”

“…The span-based model is appropriate for resolving the above problem since it completely abstains from the sequence labeling mechanism which eliminates the requirement for label dependency. Previous span-based NER models [1,[7][8][9] are proposed to recognize nested entities by regarding text spans as candidate entities where spans can be nested. Text spans are continuous text segments of which the length is restricted by a length threshold [10,11].…”