<?xml version="1.0" standalone="yes"?>
<Paper uid="C02-1158">
  <Title>Study of Practical Effectiveness for Machine Translation Using Recursive Chain-link-type Learning</Title>
  <Section position="2" start_page="0" end_page="0" type="intro">
    <SectionTitle>
1 Introduction
Rule-Based Machine Translation(MT)(Hutchins
</SectionTitle>
    <Paragraph position="0"> and Somers, 1992) requires large-scale knowledge to analyze both source language(SL) sentences and target language(TL) sentences.</Paragraph>
    <Paragraph position="1"> Moreover, it is difficult for a developer to completely describe large-scale knowledge that can analyze various linguistic phenomena. Therefore, Rule-Based MT is time-consuming and expensive. Statistical MT and Example-Based MT have been proposed to overcome the difficulties of Rule-Based MT. These approaches correspondto Corpus-Basedapproach. Corpus-Based approach uses translation examples that keep including linguistic knowledge. This means that the system can improve the quality ofitstranslationonlybyaddingnewtranslation examples. However, in Statistical MT(Brown et al., 1990), large amounts of translation examples are required in order to obtain high-quality translation. Moreover, Example-Based MT(Sato and Nagao, 1990; Watanabe and Takeda, 1998; Brown, 2001; Carl, 2001) which relies on various knowledge resources results in the same difficulties as Rule-Based MT. Therefore, Example-Based MT, which automatically acquires the translation rules from only bilingual text corpora, is very effective. However, existing Example-Based MT systems using the learning algorithms require large amounts of translation pairs to acquire high-quality translation rules.</Paragraph>
    <Paragraph position="2"> In Example-Based MT based on analogical reasoning(Malavazos, 2000; Guvenir, 1998), the different parts are replaced by variables to generalize translation examples as shown in (1) of Figure 1. However, the number of different parts of the two SL sentences must be same as the number of different parts of the two TL sentences. This means that the condition of acquisition of translation rules is very strict because this method allows only n:n mappings in the number of the different parts between the SL sentences and the TL sentences. As a result, many translationrulescannot be acquired.</Paragraph>
    <Paragraph position="3"> (McTait, 2001) generalizes both the different parts and the common parts as shown in Figure 1(2). This means that (McTait, 2001) allows m:n mappings in the number of the different parts, or the number of the common parts. However, itisdifficulttoacquirethetranslation rules that correspond to the lexicon level. On the other hand, we have proposed a method of Machine Translation using Inductive Learning with Genetic Algorithms(GA-ILMT)(Echizenya et al., 1996). This method automatically generates the similar translation examples from only given translation examples by applying genetic algorithms(Goldberg, 1989) as shown in (3a) of Figure 1. Moreover, the system performs Inductive Learning. By using Inductive Learning, the abstract translation rules are acquired by performing phased extraction of different parts as shown in Figure 1(3b) and (3c). In all methods shown in Figure 1, the condition of acquisition of translation rules is that two similar translation examples must exist. As a result, the systems require large amounts of translation examples.</Paragraph>
    <Paragraph position="4"> We propose a method of MT using Recursive Chain-link-type Learning as a method to overcome the above problem. In our method, the system acquires new translation rules from sparse data using other already acquired translation rules. For example, first, translation rule B is acquired by using translation rule A when the translation rule A exists in the dictionary. Moreover, translation rule C is acquired by using the translation rule B. Such a process of acquisition of translation rules is like a chain where each ring is linked. Therefore, we call this mechanism Recursive Chain-link-type Learning(RCL).Thismethodcaneffectively acquire many translation rules from sparse data withoutdependingon thedifferentparts ofsimilar translation pairs. In this paper, we describe the effectiveness of RCL through evaluation experiments. null</Paragraph>
  </Section>
class="xml-element"></Paper>