Translating into Free Word Order Languages

Translating into Free Word Order Languages Beryl Hoffman Centre for Cognitive Science University of Edinburgh 2 Buccleuch Place Edinburgh, EH8 9LW, U.K. hoffman~cogsci, ed. ac. uk Abstract In this paper, I discuss machine trans- lation of English text into a relatively "free" word order language, specifically Turkish. I present algorithms that use contextual information to determine what the topic and the focus of each sen- tence should be, in order to generate the contextually appropriate word orders in the target language. 1 Introduction Languages such as Catalan, Czech, Finnish, Ger- man, Hindi, Hungarian, Japanese, Polish, Rus- sian, Turkish, etc. have much freer word order than English. For example, all six permutations of a transitive sentence are grammatical in Turk- ish (although SOV is the most common). When we translate an English text into a "free" word or- der language, we are faced with a choice between many different word orders that are all syntacti- cally grammatical but are not all felicitous or con- textually appropriate. In this paper, I discuss ma- chine translation (MT) of English text into 2hrk- ish and concentrate on how to generate the appro- priate word order in the target language based on contextual information. The most comprehensive project of this type is presented in (Stys/Zemke, 1995) for MT into Pol- ish. They use the referential form and repeated mention of items in the English text in order to predict the salience of discourse entities and or- der the Polish sentence according to this salience ranking. They also rely on statistical data, choos- ing the most frequently used word orders. I argue for a more generative approach: a particular in- formation structure (IS) can be determined from the contextual information and then can be used to generate the felicitous word order. This paper concentrates on how to determine the IS from con- textual information using centering, old vs. new information, and contrastiveness. (Hajifiov£/etal, 1993; Steinberger, 1994) present approaches that determine the IS by using cues such as word order, definiteness, and complement semantic types (e.g. temporal adjuncts vs arguments) in the som:cc language, English. I believe that we cannot rely upon cues in the source language in order to de- termine the IS of the translated text. Instead, I use contextual informati<)n in the target language to determine the IS of sentences in the target lan- guage. In section 2, I discuss the Information Struc- ture, and specifically th<~ topic and the focus in naturally occurring Turkish data. Then, in section 3, I present algorithms for determining the topic and the focus, and show that we can generate con- textually appropriate word orders in '[~/rkish using these algorithms in a simple MT implementation. 2 Information Structure ]n the Information Structure (IS) that I use for Turkish, a sentence is first divided into a topic and a comment. The topic is the maiu ele- ment that the sentence is about, and the com- ment is the information conveyed about this toI)ic. Within the comment, we tind the …

Content truncated to 500 words