Nested Named Entity Recognition

Proceedings of the 2009 Conference on Empirical Methods in Natural Language Processing, pages 141–150, Singapore, 6-7 August 2009. c⃝2009 ACL and AFNLP Nested Named Entity Recognition Jenny Rose Finkel and Christopher D. Manning Computer Science Department Stanford University Stanford, CA 94305 {jrfinkel|manning}@cs.stanford.edu Abstract Many named entities contain other named entities inside them. Despite this fact, the field of named entity recognition has al- most entirely ignored nested named en- tity recognition, but due to technological, rather than ideological reasons. In this pa- per, we present a new technique for rec- ognizing nested named entities, by using a discriminative constituency parser. To train the model, we transform each sen- tence into a tree, with constituents for each named entity (and no other syntactic struc- ture). We present results on both news- paper and biomedical corpora which con- tain nested named entities. In three out of four sets of experiments, our model outperforms a standard semi-CRF on the more traditional top-level entities. At the same time, we improve the overall F-score by up to 30% over the flat model, which is unable to recover any nested entities. 1 Introduction Named entity recognition is the task of finding en- tities, such as people and organizations, in text. Frequently, entities are nested within each other, such as Bank of China and University of Wash- ington, both organizations with nested locations. Nested entities are also common in biomedical data, where different biological entities of inter- est are often composed of one another. In the GENIA corpus (Ohta et al., 2002), which is la- beled with entity types such as protein and DNA, roughly 17% of entities are embedded within an- other entity. In the AnCora corpus of Spanish and Catalan newspaper text (Mart´ı et al., 2007), nearly half of the entities are embedded. However, work on named entity recognition (NER) has almost en- tirely ignored nested entities and instead chosen to focus on the outermost entities. We believe this has largely been for practical, not ideological, reasons. Most corpus designers have chosen to skirt the issue entirely, and have annotated only the topmost entities. The widely used CoNLL (Sang and Meulder, 2003), MUC-6, and MUC-7 NER corpora, composed of American and British newswire, are all flatly annotated. The GENIA corpus contains nested entities, but the JNLPBA 2004 shared task (Collier et al., 2004), which utilized the corpus, removed all embedded entities for the evaluation. To our knowledge, the only shared task which has included nested enti- ties is the SemEval 2007 Task 9 (M´arquez et al., 2007b), which used a subset of the AnCora corpus. However, in that task all entities corresponded to particular parts of speech or noun phrases in the provided syntactic structure, and no participant di- rectly addressed the nested nature of the data. Another reason for the lack of focus on nested NER is technological. The NER task arose in the context of the MUC workshops, as small chunks which could be identified by finite state models or gazetteers. This then …

Content truncated to 500 words