eSpatialML: An Event-Driven Spatial Annotation Framework

eSpaceML: An Event-Driven Spatial Annotation Framework∗ Kiyong Lee a, Jonathan Webster b and Alex Chengyu Fang b aDepartment of Linguistics, Korea University Anam-dong, Seoul, Korea klee@korea.ac.kr bDepartment of Chinese, Translation and Linguistics, City University of Hong Kong Tat Chee Avenue, Kowloon, Hong Kong SAR {ctjjw,acfang}@cityu.edu.hk Abstract. This paper proposes eSpaceML as a representation scheme for annotating event- driven spatial expressions in natural language. It adopts SpatialML (MITRE, 2009) and ISO-Space (ISO, 2010) as a basis for the development of a novel, distributed spatial annotation scheme. SpatialML focuses on the annotation of spatial locations and their topological relations, while both ISO-Space and eSpaceML attempt to extend the scope beyond the treatment of toponyms. ISO-Space and eSpaceML also link space to events in various ways but with considerable differences. Unlike ISO-Space, which attempts to provide a self-contained framework for the various links, eSpaceML treats them in a distributed manner, operating as a pivotal system that refers to several other established annotation schemes such as MAF (ISO, 2008) and ISO-TimeML (ISO, 2009c) for morpho- syntactic as well as temporal-eventual annotations. Keywords: semantic annotation, space and event, ISO, language resource management 1 Introduction: Aim and Scope This paper aims at a description of the development of eSpaceML, an event-driven spatial annotation scheme for natural language texts. This work adopts SpatialML (MITRE, 2009) as a basic framework for annotating spatial locations and their topological relations. The scope of eSpaceML, however, goes beyond the annotation of place names and their relations and attempts to extend its task to events by linking events to space, as is implied by the prefix e- in its name eSpaceML.1 The idea of linking events to space has originated from ISO-Space (ISO, 2010). At its current preliminary stage, ISO-Space annotates events as well as space, and also signals for linking those two directly in a self-contained manner. Our proposed eSpaceML, on the other hand, is designed to outsource various tasks other than the main task of linking events to space. For the annotation of events, for instance, it refers to ISO-TimeML (ISO, 2009c) that annotates events and time. For the reference to markables, i.e., those expressions that need to be marked up, it uses the information provided by other annotation frameworks such as MAF (ISO, 2008), which ∗ The work reported in this paper was supported in part by research grants from the City University of Hong Kong (Project Nos 9610053, 9610126, 7008002, 7002387 and 7002190). The authors would like to thank Roland Hausser and the reviewers for their valuable comments and suggestions. Copyright 2010 by Kiyong Lee, Jonathan Webster and Alex Chengyu Fang 1 As defined in ISO-TimeML (ISO, 2009c), the term event covers all of the types of eventualities such as states, processes, and transitions. See also Pustejovsky (1991) for these three event types. PACLIC 24 Proceedings 223 provides morpho-syntactic information. The proposed eSpaceML is thus designed to provide a net-based distributed annotation scheme in a pivotal format. This paper claims that eSpaceML admits a compositional approach to semantic …

Content truncated to 500 words