Categorial divergences

Abstract

The topic of this chapter will be the class of translation problems which are called categorial divergences as introduced in Chapter 12. Given the controlled M-grammar framework introduced in Chapter 9, the most obvious way to attune grammars is to attune the corresponding subgrammars of the different grammars per category to each other. The consequence of this straightforward tuning is that words and phrases of a certain syntactic category are always translated into words and phrases of the same category. Thus, isomorphy between BVtoCLAUSE subgrammars assumes that a verb is always translated into a verb and a clause into a clause. Isomorphy between BNtoNP subgrammars assumes that a noun is always translated into a noun and an NP into an NP. This may be preferable, but is not always possible, due to categorial divergences between languages. Some examples of this type of divergence are shown in Table 13.1. In other MT systems, structural transfer (cf. Estival et al. (1990)) or complex lexical transfer. (cf. Thurmair (1990)) is required for these divergences. This chapter will be organised in the following way: in Section 13.2 two types of categorial divergences will be distinguished: head category divergences and top category divergences. Section 13.3 will introduce the technique of monolingual isomorphic subgrammars, which offers systematic solutions for both types. Section 13.4 discusses a further subdivision of projection subgrammars and Section 13.5 the issue of complete and partial isomorphy of subgrammars. Section 13.6 shows how multiple divergences within one clause can be dealt with. In Section 13.7 a few alternative approaches will be discussed, followed by some concluding remarks in Section 13.8.