A Neural Layered Model for Nested Named Entity Recognition

Proceedings of NAACL-HLT 2018, pages 1446–1459 New Orleans, Louisiana, June 1 - 6, 2018. c⃝2018 Association for Computational Linguistics A Neural Layered Model for Nested Named Entity Recognition Meizhi Ju1,3, Makoto Miwa2,3 and Sophia Ananiadou1,3 1National Centre for Text Mining, University of Manchester, United Kingdom 2Toyota Technological Institute, Japan 3Artificial Intelligence Research Center (AIRC), National Institute of Advanced Industrial Science and Technology (AIST), Japan {meizhi.ju, sophia.ananiadou}@manchester.ac.uk makoto-miwa@toyota-ti.ac.jp Abstract Entity mentions embedded in longer entity mentions are referred to as nested entities. Most named entity recognition (NER) sys- tems deal only with the flat entities and ignore the inner nested ones, which fails to capture finer-grained semantic information in underly- ing texts. To address this issue, we propose a novel neural model to identify nested enti- ties by dynamically stacking flat NER layers. Each flat NER layer is based on the state-of- the-art flat NER model that captures sequen- tial context representation with bidirectional long short-term memory (LSTM) layer and feeds it to the cascaded CRF layer. Our model merges the output of the LSTM layer in the current flat NER layer to build new represen- tation for detected entities and subsequently feeds them into the next flat NER layer. This allows our model to extract outer entities by taking full advantage of information encoded in their corresponding inner entities, in an inside-to-outside way. Our model dynamically stacks the flat NER layers until no outer enti- ties are extracted. Extensive evaluation shows that our dynamic model outperforms state-of- the-art feature-based systems on nested NER, achieving 74.7% and 72.2% on GENIA and ACE2005 datasets, respectively, in terms of F- score.1 1 Introduction The task of named entity recognition (NER) in- volves the extraction from text of names of en- tities pertaining to semantic types such as per- son (PER), location (LOC) and geo-political en- tity (GPE). NER has drawn the attention of many researchers as the first step towards NLP applica- tions such as entity linking (Gupta et al., 2017), re- lation extraction (Miwa and Bansal, 2016), event 1Code is available at https://github.com/ meizhiju/layered-bilstm-crf LOC PER GPE GPE The premier of the western Canadian province of British Columbia ... LOC PER GPE GPE The premier of the western Canadian province of British Columbia ... Figure 1: A sentence from ACE2005 (Walker et al., 2006) containing the nested 4 entities nested 3 levels deep. extraction (Feng et al., 2016) and co-reference res- olution (Fragkou, 2017; Stone and Arora, 2017). Due to the properties of natural language, many named entities contain nested entities: embedded names which are included in other entities, illus- trated in Figure 1. This phenomenon is quite com- mon in many domains (Alex et al., 2007; Byrne, 2007; Wang, 2009; M`arquez et al., 2007). How- ever, much of the work on NER copes only with non-nested entities which are also called flat enti- ties and neglects nested entities. This leads to loss of potentially important information, with nega- tive impacts on subsequent tasks. Traditional approaches to NER mainly in- volve two …

Content truncated to 500 words