Automatic translation of noun compounds

Abstract

This paper describes the treatment of nominal compounds in a tranarer based machine translation system; it presents a new approach for resolving ambiguities in compound segmentation and Constituent structure selection using a combination of linguistic rules and statistical data. An introduction to the general as well as to the German-English-specific problems of compound translation is given (sect. 1). In section 2, the analysis phase is described with its linguistics as well as its computational aspects. Section 3 deals with the transfer and generation process, focussing on corpus based techniques.