Why Is Zero Marking Important in Korean?

R. Dale et al. (Eds.): IJCNLP 2005, LNAI 3651, pp. 588 – 599, 2005. © Springer-Verlag Berlin Heidelberg 2005 Why Is Zero Marking Important in Korean? Sun-Hee Lee1,*, Donna K. Byron2, and Seok Bae Jang3 1,2 395, Dreese Lab., 2015, Neil Avenue,Columbus, OH 43210 shlee@ling.ohio-state edu dbyron@cse.ohio-state.edu 3 37th and O Sts., NW, Washington, D.C, 20057 sbj3@georgetown.edu Abstract. This paper argues for the necessity of zero pronoun annotations in Korean treebanks and provides an annotation scheme that can be used to de- velop a gold standard for testing different anaphor resolution algorithms. Rele- vant issues of pronoun annotation will be discussed by comparing the Penn Ko- rean Treebank with zero pronoun mark-up and the newly developing Sejong Teebank without zero pronoun mark-up. In addition to supportive evidence for zero marking, necessary morphosyntactic and semantic features will be sug- gested for zero annotation in Korean treebanks. 1 Introduction This paper discusses the importance of zero pronoun marking in treebanks and investi- gates what kind of linguistic features are needed for treebank annotation in order to increase the usability of annotated corpora. Zero pronouns refer to empty pronouns without phonological realization, which work in a similar manner as English pronouns. In the recent decade, there has been remarkable progress in the realm of building large corpora in Korean and applying them for linguistic research and natural lan- guage processing. Based on the broad acknowledgement of the importance of corpus and applicative tools, the 21st century Sejong project was launched in 1998 and has been developing various database and relevant computational tools including elec- tronic dictionaries, annotation tools, morphological analyzers, parsers, etc. As a part of the Sejong project, the syntactically annotated treebank of Korean has been under construction. In addition to the Sejong Treebank (henceforth, ST), the Penn Korean Treebank (Han et al.[4] henceforth, PKT) has already been released and continues to be expanded. These treebanks with abundant linguistic information are expected to fulfill a function as informative databases in broad domains of theoretical linguistics and computational linguistics such as statistical approaches, machine learning, etc. A notable point is that there is a critical difference between annotations of ST and PKT with respect to marking zero elements including traces, zero pronouns, etc. The most current guidelines of ST specify that zeros are dropped in order to maintain the * This work was supported by the Korea Research Foundation Grant (KRF-2004-037-A00098) for the author. Why Is Zero Marking Important in Korean? 589 consistency and efficiency of the treebank. In contrast, PKT advocates for representing zero elements. According to different approaches to zero marking, the structure of the following sentence (1a) is differently analyzed as in (1b) and (1c); in (1b) ST does not contain any missing subject while PKT marks the missing subject and object as pros in (1c)1 (1) a. 어제밤 12시-에 받-았-습니다. eceypam 12 si-ey pat-ass-supnita. last night 12 o’clock-at receive-Past-E ‘Last night at 12 o’clock, (I/(s)he/they) received (it)’. b. ST: (VP (AP 어젯밤/MAG) (VP (NP_AJT 12/SN + 시/NNB + 에/JKB) …

Content truncated to 500 words