Abstract:
One of the internal rules of the example-based translation is the dependence of the
translation quality on the length and representation of translation examples. These are managed
by an example database, for which the linguist should answer two questions when s/he intends
to design it: (1) what is the representation chosen for the translation example? and (2) what are
the possibilities to generalize the translation examples stored in the database? In this paper my
intent is to propose two answers, in fact – a synergetical research direction for Romanian,
English and French language. I will consider the representation of the translation examples as
depedency trees and I will generalize them by means of the semantic information described by
Levin’s verb classes.
