A Deterministic Method for Structural Analysis of Compound Words in Japanese

A Deterministic Method for Structural Analysis of Compound Words in Japanese Dongli Han, Takeshi Ito and Teiji Furugori Department of Computer Science, The University of Electro-Communications 1-5-1 Chofugaoka, Chofu-shi, Tokyo 182-8585, Japan {han, ito, furugori@phaeton.cs.uec.ac.jp} Abstract Structural analysis of compound words is necessary and an important process in natural lan- guage processing. Proposed here is a corpus- and statistics- based method for the structural analysis of compound words in Japanese. We determine the structure of a compound word by using Internet corpus and calculating the strength of word association among its constitu- ent words. Experiments with 5, 6, 7, and 8 kanji compound words show that our method works well and its performance is better than those of other comparable studies. 1� Introduction A sequence of words may assume a single syntactic function in English and the same is true in Japa- nese. Machine translation system, for instance, consists of three words machine, translation and system. Its equivalent in Japanese, MA FIR 3 5-1A, is a compound word with three constituent words, OW (machine), (translation), and (system). It is a single concept and functions as a noun. Both English and Japanese face problems in processing such a sequence of words. A problem involved in English is in identifying the sequence of words as a syntactic as well as semantic unit. It is easy to find a compound word in Japanese but we have difficulties in segmenting it into con- stituent words as well as determining its syntactic or semantic structure among the constituents. Finding its constituents and determining the structure of a compound word is an important proc- ess in constructing practical natural language systems for machine translation, information retrieval, text summarization, etc. For the compound word fc H gait (the former president of Bank of Japan), for instance, there are six lexically possible segmentations: 7-E/ H I iei g, 7GI 19 /, /en, H /*tit&� H /WA, 7G H /e/It and 5---C H / ;F /glia, the correct segmentation here being 7c(former)/ H (Bank of Japan)/a ot (president). Standing alone, the word segmentation is useful in many areas of natural language processing. However, we should know how the constituents in a compound word are combined to make the meaning, say, for its translation into foreign languages. 7G/19 "d /km has two probable depend- ency structures: (CAM H )(a4))) and (((A)( H ))(M)). The former leads to the meaning of the former president of Bank of Japan and the latter the president of former Bank of Japan. The problem is how can we arrive at the structure, ((i-E)(( H 1A1F)(atit))). In this paper we offer a method for analyzing the structure of compound words in Japanese. In Section 2 describes some previous work done for a compound word on the identification of con- stituents and determination of its structure. We propose a method and subsequently an algorithm in 79 Section 3 for analyzing the structure of compound words using the strength of association measures among the constituent words. In Section …

Content truncated to 500 words