A Chinese POS Decision Method Using Korean Translation Information

In this paper we propose a method that imitates a translation expert using the Korean translation information and analyse the performance. Korean is good at tagging than Chinese, so we can use this property in Chinese POS tagging. Keyword : machine translation, part of speech tagging, corpus Introduction Previous POS(Part Of Speech) tagging methods of Chinese can be largely classified into 2. One is a method using POS tagging rules that is extracted by Chinese experts, and the other is to u se a statistical model, it usually needs huge Chinese corpus. In case of use the first method, it is very difficult to extract rules and these rules can’t represent va rious language examples. So this method was used in the early years of POS tagging and has been used together with other methods. Other method, corpus-based statistical methods have many advantages, but it needs large amount of POS corpus. To solve these problems, some researchers are studying about a method that uses raw text data ext racted from internet[2, 5, 6]. In this paper we propose a POS decision method imitate Chinese translator whose native language is Korean and POS decision rule extracting method using Chinese –Korean bilingual corpus. 1. Availability of Korean translation information in Chinese POS decision Machine translation system is a typical expert system, so the closer translation method to translati on expert, the higher accuracy of the translation system. Translation expert whose mother tongue is Korean generate Korean words from Chinese words w hen he translates Chinese, and verifies their translation and finally complete the translation. For example, consider following sentence. 精密的观察是科学研究的基础(세밀한 관찰은 과학연구의 기초이다.) When the expert translates this sentence, expert generate Chinese words ‘科学研究’ from 《과학 연구》, and he knows these words are right words and 《과학은 연구한다》are not right through his experience. Classical method is to divide the sentence as 精密/的/观察/是/科学/研究/的/基础, and decide the POS of multi-POS word‘研究’using statistical information. This method largely depends on the amount of corpus and their fields. We can use Korean POS tagging system for this problem. Using the POS tagged words of 《세밀한 관찰은 과학연구의 기초이다.》, we can realize the n oun combination of 《과학연구》and it tells us the POS of Chinese word ‘研究’is noun. In conclusion using large amount of Korean POS tagged corpus, we can improve the performance of Chinese POS tagging system. 2. Acquisition method of POS decision rule from Chinese-Korean bilingual corpus. Above-mentioned method can be applied into noun noun conjunction, but it can not be applied in to verb decision and so on. Generally we use following rules for the POS decision of Chinese multi POS words. setpos(0,v) Acquisition of these rules generally needs many efforts and it is difficult to decide the confidence of rule. We can use Chinese-Korean bilingual corpus to extract statistical rule. Consider the acquisition method through a example. Chinese sentence: 我的朋友学习中国语 Translation: 나의 동무는 중국어를 배운다 For above example, do the word dividing for Chinese and do the POS tagging for Korean. 我/的/朋友/学习/中国语 나/N 의/T 동무/N 는/T 중국어/N 를/T 배우다/V ᄂ다/T In this sentence, morpheme analysis of 《배운다》is 《배우다/동사 + ᄂ다/맺음토》and verb t ranslation of 学习 is 《배우다》, so 学习 is used in verb. Chinese sentence Cs and Korean sentence Ks can be represented as follows.