NNE: A Dataset for Nested Named Entity Recognition in English Newswire

Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics, pages 5176–5181 Florence, Italy, July 28 - August 2, 2019. c⃝2019 Association for Computational Linguistics 5176 NNE: A Dataset for Nested Named Entity Recognition in English Newswire Nicky Ringland1 † Xiang Dai1,2 ‡ Ben Hachey1,3 Sarvnaz Karimi2 Cecile Paris2 James R. Curran1 1University of Sydney, Sydney, Australia 2CSIRO Data61, Sydney, Australia 3Digital Health CRC, Sydney, Australia †nicky.ringland@sydney.edu.au ‡dai.dai@csiro.au Abstract Named entity recognition (NER) is widely used in natural language processing applica- tions and downstream tasks. However, most NER tools target flat annotation from popu- lar datasets, eschewing the semantic informa- tion available in nested entity mentions. We describe NNE—a fine-grained, nested named entity dataset over the full Wall Street Journal portion of the Penn Treebank (PTB). Our an- notation comprises 279,795 mentions of 114 entity types with up to 6 layers of nesting. We hope the public release of this large dataset for English newswire will encourage development of new techniques for nested NER. 1 Introduction Named entity recognition—the task of identifying and classifying entity mentions in text—plays a crucial role in understanding natural language. It is used for many downstream language process- ing tasks, e.g., coreference resolution, question answering, summarization, entity linking, relation extraction and knowledge base population. How- ever, most NER tools are designed to capture flat mention structure over coarse entity type schemas, reflecting the available annotated datasets. Focusing on flat mention structures ignores im- portant information that can be useful for down- stream tasks. Figure 1 includes examples of nested named entities illustrating several phenomena: • Entity-entity relationships can be embedded in nested mentions. For instance, the location of the ‘Ontario Supreme Court’ is indicated by the embedded STATE mention ‘Ontario’; • Entity attribute values can be embedded in nested mentions. For instance, the title is the embedded ROLE ‘Former U.N. Ambassador’, which also encodes the employment relation ... the Ontario Supreme Court said it will postpone ... state government Former U.N. Ambassador Jeane Kirkpatrick ... org:other role first name role per role per ... this wealthy Southern California beach community ... state region Figure 1: Example nested mentions in NNE. between the PERSON ‘Jane Kirkpatrick‘ and ORG ‘U.N.’; • Part-whole relationships can be encoded in nested mention structure. For instance, the REGION ‘Southern California’ is part of the STATE ‘California’. Recent work has demonstrated increasing in- terest in nested entity structure, including lo- cal approaches (Xu et al., 2017; Sohrab and Miwa, 2018), hypergraph-based approaches (Lu and Roth, 2015; Muis and Lu, 2017; Katiyar and Cardie, 2018; Wang and Lu, 2018), cascaded ap- proaches (Alex et al., 2007; Ju et al., 2018), and parsing approaches (Finkel and Manning, 2009; Wang et al., 2018). See Dai (2018) for a survey. Yet these techniques have seen little translation from the research literature to toolsets or down- stream applications. To facilitate ongoing research on nested NER, we introduce NNE—a large, manually- annotated, nested named entity dataset over En- glish newswire. This new annotation layer over the …

Content truncated to 500 words