Learning to Recognize Discontiguous Entities

Proceedings of the 2016 Conference on Empirical Methods in Natural Language Processing, pages 75–84, Austin, Texas, November 1-5, 2016. c⃝2016 Association for Computational Linguistics Learning to Recognize Discontiguous Entities Aldrian Obaja Muis Wei Lu Singapore University of Technology and Design {aldrian muis,luwei}@sutd.edu.sg Abstract This paper focuses on the study of recognizing discontiguous entities. Motivated by a previ- ous work, we propose to use a novel hyper- graph representation to jointly encode discon- tiguous entities of unbounded length, which can overlap with one another. To compare with existing approaches, we first formally in- troduce the notion of model ambiguity, which defines the difficulty level of interpreting the outputs of a model, and then formally analyze the theoretical advantages of our model over previous existing approaches based on linear- chain CRFs. Our empirical results also show that our model is able to achieve significantly better results when evaluated on standard data with many discontiguous entities. 1 Introduction Building effective automatic named entity recogni- tion (NER) systems that is capable of extracting useful semantic shallow information from texts has been one of the most important tasks in the field of natural language processing. An effective NER sys- tem can typically play an important role in certain downstream NLP tasks such as relation extraction, event extraction, and knowledge base construction (Hasegawa et al., 2004; Al-Rfou and Skiena, 2012). Most traditional NER systems are capable of ex- tracting entities1 as short spans of texts. Two ba- sic assumptions are typically made when extract- 1Or sometimes mentions are considered, which can be named, nominal or pronominal references to entities (Florian et al., 2004). In this paper we use “mentions” and “entities” interchangeably. EGD showed [hiatal hernia]1 and vertical [laceration]2 in distal [esophagus]2 with [blood in [stomach]4]3 and overlying [lac]4. Figure 1: Discontiguous entities in a medical domain. Words annotated with the same index are part of the same entity. Note that entity 3 and entity 4 overlap with one another. ing entities: 1) entities do not overlap with one an- other, and 2) each entity consists of a contiguous se- quence of words. These assumptions allow the task to be modeled as a sequence labeling task, for which many existing models are readily available, such as linear-chain CRFs (McCallum and Li, 2003). While the above two assumptions are valid for most cases, they are not always true. For example, in the entity University of New Hampshire of type ORG there exists another entity New Hampshire of type LOC. This violates the first assumption above, yet it is crucial to extract both entities for subsequent tasks such as relation extraction and knowledge base construction. Researchers therefore have proposed to tackle the above issues in NER using more so- phisticated models (Finkel and Manning, 2009; Lu and Roth, 2015). Such efforts still largely rely on the second assumption. Unfortunately, the second assumption is also not always true in practice. There are also cases where the entities are composed of multiple discontiguous sequences of words, such as in disorder …

Content truncated to 500 words