KIT REPORT 106 Defaults in Machine Translation

In this paper we try to combine experiences gained in the Machine Translation project with research concerning the default extension of the terminological representation system BACK. We analyze various types of ambiguities that are problematic for translation and show that disambiguation is possible on the basis of default assumptions. Some of these defaults form preference rule systems, while others are merely an elegant way of modeling exceptions. We show how the ideas underlying the implementation can be reformulated in a declarative framework by using terminological logics. We present an approach towards disambiguation which we call interpretation as exception minimization. The basic idea is to compare the possible interpretations with respect to the defaults (preference rules) they violate. Given a relevance ordering between multisets of defaults, we chose the interpretation yielding the minimal number of exceptions with respect to this ordering as the preferred one.