Zero Object Resolution in Korean

Zero Object Resolution in Korean Arum Park Seunghee Lim Munpyo Hong Dept. of German Linguistics & Literature, Sungkyunkwan University / 25-2, Sungkyunkwan-Ro, Jongno-Gu, Seoul, Korea Dept. of German Linguistics & Literature, Sungkyunkwan University / 25-2, Sungkyunkwan-Ro, Jongno-Gu, Seoul, Korea Dept. of German Linguistics & Literature, Sungkyunkwan University / 25-2, Sungkyunkwan-Ro, Jongno-Gu, Seoul, Korea remin2@skku.edu rusilen21@skku.edu skkhmp@skku.edu Abstract Korean is one of the well-known „pro-drop‟ languages. When translating Korean zero object into languages in which objects have to be overtly expressed, the resolution of zero object is crucial. This paper proposes a machine learning method to resolve Korean zero object. We proposed 8 linguistically motivated features for ML (Machine Learning). Our approach has been implemented with WEKA 3.6.10 and evaluated by using 10-fold cross validation method. The accuracy of the proposed method reached 73.37%. 1 Introduction Korean is one of the so-called pro-drop languages. Certain pronouns may be omitted and the omitted pronouns are often called zero pronouns. This kind of pronoun also occurs in other languages, such as Japanese or Spanish. The omitted pronouns in Korean can appear in subject and object position, whereas in Spanish or Italian, they can appear only in subject position. A zero subject is the most frequent type of anaphoric expressions in Korean. Hong (2000) reported that about 57% of the pronouns are a zero subject pronoun in pronoun occurrences in Korean spoken text.  Corresponding author Zero object is the second most frequent zero pronoun type in Korean spoken text. Despite of the frequent use of zero objects, most of the previous works do not deal with the zero objects in Korean. In this work, we focus on Korean spoken texts, since zero pronouns occur more frequently than in a written text. Ryu (2001) showed that a zero pronoun rarely appears in written texts when it is compared with spoken texts in Korean. For this reason, we conduct a study for Korean zero object in Korean spoken text and try to find the linguistic clues for the zero object resolution. In the context of machine translation, the resolution of Korean zero objects could be one of the most important issues in order to translate them into the target language like English and German. One of the reasons that zero objects in Korean is a problem in MT is that the omitted objects in Korean have to be translated into overt objects in target languages. Unfortunately, the majority of MT systems do not deal with this problem, because most of the current commercial MT systems do not treat the linguistic phenomena that go beyond a sentence level. To illustrate this issue, let's take a look at the following example (1). (1) MT results from Korean(a) to German(b) (a) Korean A: 여권 i 을 분실했습니다. PACLIC 29 439 29th Pacific Asia Conference on Language, Information and Computation pages 439 - 448 Shanghai, China, October 30 - November 1, 2015 Copyright 2015 by Arum Park, Seunghee Lim and Munpyo Hong a passport-OBJ lost (yekwenul punsilhaysssupnita.) “I lost …

Content truncated to 500 words