Translation Template Learning Based on Hidden Markov Modeling

Translation Template Learning Based on Hidden Markov Modeling Nguyen Minh Le Graduate School of Information Science, JAIST ISHIKAWA 923-1292,JAPAN nguyenml@jaist.ac . jp Akari Shimazu Graduate School of Information Science, JAIST ISHIKAWA 923-1292, JAPAN shimazu@jaist.ac.jp Susumu Horiguchi Graduate School of Information Science, JAIST ISHIKAWA 923-1292, JAPAN hori@jaist.ac.jp Abstract This paper addresses a novel translation method based on Hidden Markov Model using template rules after learning them from the bilingual corpus. The method can enhance the translation accuracy and ensure a low complexity in comparing with the pervious template learning translation method and draws a new perspective for applying statistical machine learning on example based translations. domain. Keywords: Machine translation, EBMT, Template learning translation, HMM. 1 Introduction Example based machine translation (EBMT), originally proposed by Nagao (Nagao,M.A 1984), and is one of the main approaches to corpus-based machine translation. The main idea behind EBMT is that an input sentence in the source language is compared with the example translations in the given bilingual parallel text to find the closest matching examples so that these examples can be used in the translation of the input sentence. After finding the closest matches for the sentence in the source language, parts of the corresponding target language sentence are constructed using structural equivalences and deviances in the matches. Following Nagao's original proposal, several approaches using the example based method were presented. One of the approaches that applied the idea for translation from English to Turkey is learning translation template (Cicekli,I 1996) (Gilvenir,H.A 1998). This method relies on the technique that uses the similarity and difference from a source sentence and a target sentence in the given bilingual corpus to build template rules for translation. The advantage of this method is that does not need any complex pars- ing such as syntactic parsing or semantic parsing and overcome the imperfectness of the rule-based ma- chine translation. One of the disadvantages of the method is that a lot of templates can be matched with an input sentence. To overcome this problem, (Oz and Cicekli,I 1998) present a method which allows sorting template rules according to their confident factors. The translation results are sorted using its score through the value of confident factors. However, this method needs to evaluate all matching rules for each input sentence to obtain the output results, while much of them are redundant rules. The exponential cal- culation problem will arise when an input sentence is long and the number of template rules is large. Fol- lowing that point, we present a novel method based on an HMM model that uses constraints for set of matching rules with each input sentence. Thus, the translation results of an input sentence are obtained by finding a set of template rules that is most likely with our HMM model. The remainder of this paper is organized as follows: A template learning algorithm is given in Section 2. Section 3 describes a HMM modeling for translation using template rules. Section 4 show experiments on English Vietnamese translation system and …

Content truncated to 500 words