Tense and Aspect Information in a Fudr-based German French Machine Translation System

Normally, the tense and aspect systems of natural languages do not coincide and the correct translation of the tense forms needs a semantic analysis of the text or sentence which uses these forms. The problem with semantic representations is that, often, ambiguities cannot be resolved, though a remarkable effort of inferencing has been invested to the effect that the analysis of the source text is costly, nevertheless coming up with a large number of alternative representations. In this paper, we will describe how the German-French Machine Translation, system which is currently been developed at linguatec-E & S uses lexical Aktionsart information, tenseand background information in order to determine the parameters of the specific perspective under which the context perceives the new eventuality and incorporates it; we will describe how the system uses these parameters for choosing a suitable target (tense) form. In order to circumvent the complexity problem, this computation of the relevant semantic decision criteria is not built upon an explicit (deep) semantic representation, but relates to a projection of the syntactic analysis of the source sentence which is called dependence structure. This structure defines the level of transfer. It is unique with respect to the underlying syntactic analysis and it can be interpreted as a flat underspecified semantic representation (FUDR). The tense parameters rely heavily on the investigations of Hans Kamp and Christian Rohrer about the French tense system especially and the formal representation of the meaning of the tenses within discourse representation theory.