Text Structure Aiming at Machine Translation

Machine translation relies on the existence of a meaning representation which must be able to capture the semantic content of the original text. The work presented here is concerned with the automatic generation of such a representation, to be used by a machine translation system. We have used as source text scientific abstracts in Portuguese. The emphasis of the work is on the determination of the text structure of such abstracts, making use of the notions of cohesion and coherence. The main cohesion phenomena considered here is definite anaphora.

1995