Korean TimeML and Korean TimeBank

Korean TimeML and Korean TimeBank Young-Seob Jeong∗, Won-Tae Joo†, Hyun-Woo Do†, Chae-Gyun Lim†, Key-Sun Choi†, and Ho-Jin Choi† ∗NAVER LABS, NAVER Corp. Seongnam, 463-867, South Korea pinode.waider@navercorp.com †Korea Advanced Institute of Science and Technology 291 Daehak-ro, Yuseong-gu, Daejeon 305-701, South Korea {izazu1725,realstorm103,rayote,kschoi,hojinc}@kaist.ac.kr Abstract Many emerging documents usually contain temporal information. Because the temporal information is useful for various applications, it became important to develop a system of extracting the temporal information from the documents. Before developing the system, it first necessary to define or design the structure of temporal information. In other words, it is necessary to design a language which defines how to annotate the temporal information. There have been some studies about the annotation languages, but most of them was applicable to only a specific target language (e.g., English). Thus, it is necessary to design an individual annotation language for each language. In this paper, we propose a revised version of Koreain Time Mark-up Language (K-TimeML), and also introduce a dataset, named Korean TimeBank, that is constructed basd on the K-TimeML. We believe that the new K-TimeML and Korean TimeBank will be used in many further researches about extraction of temporal information. Keywords: Korean TimeML, Temporal information, Annotation language, Korean TimeBank 1. Introduction Due to the exponentially increasing number of documents available on the Web and from other sources, it has be- come important to develop methods to automatically ex- tract knowledge from unstructured, natural language docu- ments. The knowledge extracted as such is useful for vari- ous applications in the areas of information retrieval (IR), trend analysis (TA), and question answering (QA) sys- tems. Among the many aspects of extracting knowledge from documents, the extraction of temporal information has recently drawn attention. There are two well-known an- notation languages of temporal information, Time Mark- up Language (TimeML) (Pustejovsky et al., 2003) and ISO-TimeML. Although these annotation languages define many tags and attributes for representing various types of temporal information, they do not incorporate language di- versity. For example, they assume that annotation is per- formed in the token level. However, Korean is an aggluti- native language whose words are formed by joining mor- phemes together, so it can not be annotated properly in the token level. As an annotation language for Korean, the Korean TimeML (KTimeML) was proposed (Im et al., 2009), and its con- tributions can be summarized as follows: (1) it employs a morpheme-level standoff annotation scheme, (2) it takes a surface-based annotation scheme, (3) it suggests to cancel the head-only markup policy of TimeML, (4) it addresses several Korean-specific issues (e.g., the usage of signal tag for only temporal connectives), and (5) it introduces the TARSQI Toolkit for the annotation process following the KTimeML. In this paper, we argue that the KTimeML has some limitations, and propose a revised version of the KTimeML. For example, the previous KTimeML did not consider some charactersitics of Korean (e.g., a lunar calen- dar), and the morpheme-level annotation of the KTimeML makes it difficult to share the dataset. …

Content truncated to 500 words