Korean to English Translation Using Synchronous TAGs

Korean to English Translation Using Synchronous TAGs D. Egedi, M. Palmer, H.S. Park, A.K. Joshi School of Computer Information Science and Institute for Research in Cognitive Science University of Pennsylvania Philadelphia, PA 19104-6228, USA {egedi, mpalmer, hspark, joshi}@linc.cis.upenn.edu Abstract It is often argued that accurate machine translation requires reference to contextual knowledge for the correct treatment of linguistic phenomena such as dropped arguments and accurate lexical selection. One of the historical arguments in favor of the interlingua approach has been that, since it revolves around a deep semantic representation, it is better able to handle the types of linguistic phenomena that are seen as requiring a knowledge-based approach. In this paper we present an alternative approach, exemplified by a prototype system for machine translation of English and Korean which is implemented in Synchronous TAGs. This approach is essentially transfer based, and uses semantic feature unification for accurate lexical selection of polysemous verbs. The same semantic features, when combined with a discourse model which stores previously mentioned entities, can also be used for the recovery of topicalized arguments. In this paper we concentrate on the translation of Korean to English. 1 Introduction It is often argued that accurate machine translation requires reference to contextual knowledge for the correct treatment of linguistic phenomena such as dropped arguments and accurate lexical selection [3,12]. This is still, in many cases, an unsolved problem for natural language analysis [14], which adds to the burden of the already beleaguered machine translation systems. One of the historical arguments in favor of the interlingua approach has been that, since it revolves around a deep semantic representation, it is better able to handle the types of linguistic phenomena that are seen as requiring a knowledge-based approach. However, recent implementations of machine translation systems are blurring the distinction between transfer systems and interlingua systems [20,13]. In this paper we present a prototype system for machine translation between English and Korean which is implemented in the Synchronous Tree Adjoining Grammar (STAG) formalism [16], an extension of Lexicalized [15], Feature-Based [19] Tree Adjoining Grammars (FB-LTAGs) [9]. We illustrate the syntactic coverage of the system with relative clauses, wh-questions and constituent order divergence examples. Although we describe what is essentially a transfer based approach, the system uses feature unification for lexical selection and is being augmented with a discourse model to handle discourse related phenomena such as recovery of topicalized arguments. In this paper we concentrate on the translation of Korean to English. 48 2 FB-LTAG and STAG Formalism 2.1 An Introduction to FB-LTAGs The Feature-Based Tree Adjoining Grammar formalism (FB-LTAG) is based on the Tree Adjoining Grammar (TAG) formalism developed by Joshi, Levy, and Takahashi [9], which has been extended to include lexicalization [15], and unification-based feature structures [19]. As first shown by Joshi and Kroch [8,11], the properties of TAGs permit us to encapsulate diverse syntactic phenomena such as unbounded dependencies in a very natural way. The primitive elements of the standard TAG formalism are known as elementary trees. Elementary trees …

Content truncated to 500 words