Interlingual machine translation A parameterized approach

摘要

The task of designing a machine translation system is difficult because, in order to achieve any degree of accuracy, such systems must capture language-independent information while still systematically processing many types of language-specific phenomena in each of the individual languages. This paper provides a catalog of certain types of distinctions among Spanish, English, and German, and describes a parameterized interlingual approach that characterizes these distinctions, both at the syntactic level and at the lexical-semantic level. The parameter-setting approach to machine translation is desirable because it simplifies the descriptions of natural grammars, facilitates the task of modifying and augmenting the system, accounts for cross-linguistic variation uniformly, and provides a more constrained theory of processing. The approach described here is implemented in a system called UNITRAN, an interlingual machine translation system that translates English, Spanish, and German bidirectionally.