Handling Syntactical Ambiguity in Machine Translation

HANDLING SYNTACTICAL AMBIGUITY IN MACHINE TRANSLATION Vladimir Pericliev Institute of Industrial Cybernetics and Robotics Acad. O.Bontchev Sir., bl.12 1113 Sofia, Bulgaria ABSTRACT The difficulties to be met with the resolu- tion of syntactical ambiguity in MT can be at least partially overcome by means of preserving the syntactical ambiguity of the source language into the target language. An extensive study of the co- rrespondences between the syntactically ambiguous structures in English and Bulgarian has provided a solid empirical basis in favor of such an approach. Similar results could be expected for other suffi- ciently related languages as well. The paper con- centrates on the linguistic grounds for adopting the approach proposed. 1. INTRODUCTION Syntactical amblgulty, as part of the ambigui- ty problem in general, is widely recognized as a major difficulty in MT. To solve this problem, the efforts of computational linguists have been main- ly directed to the process of analysis: a unique analysis is searched (semantical and/or world knowledge information being basically employed to this end), and only having obtained such an ana- lysis, it is proceeded to the process of synthesis. On this approach, in addition to the well known difficulties of general-linguistic and computa- tional character, there are two principle embarras- ments to he encountered. It makes us entirely in- capable to process, first, sentences with "unre- solvable syntactical ambiguity" (with respect to the disambiguation information stored), and, se- condly, sentences which must he translated ambi- guously (e.g. puns and the like). In this paper, the burden of solution of the syntactical ambiguity problem is shifted from the domain of analysis to the domain of synthesis of sentences. Thus, instead of trying to resolve such ambiguities in the source language (SL), syntac- tically ambiguous sentences are synthesized in the target language (TL) which preserve their ambigui- ty, so that the user himself rather than the par- ser disambiguates the ambiguities in question. This way of handling syntactical ambiguity may be viewed as an illustration of a more gene- ral approach, outlined earlier (Penchev and Perl- cliev 1982, Pericliev 1983, Penchev and Perlcllev 1984), concerned also with other types of ambt- guitles in the SL translated by means of syntacti- cal, and not only syntactical, ambiguity in the TL. In this paper, we will concentrate on the linguistics ~rounds for adopting such a manner of handling of syntactical ambiguity in an English in- to Bulgarian translation system. 2. PHILOSOPHY This approach may be viewed as an attempt to simulate the behavior of s man-translator who is linguistically very competent, but is quite unfa- miliar with the domain he is translating his texts from. Such a man-translator will be able to say what words in the original and in the translated sentence go together under all of the syntactica- lly admissible analyses; however, he will be, in general, unable to make a decision as to which of these parses "make sense". Our approach will be an obvious way out of this situation. And it is in fact not …

Content truncated to 500 words