The discussion in previous chapters reinforces the point made in
Chapter 
about the value of syntactic, and `shallow'
semantic analysis, but it also shows why performing a syntactic
analysis alone is not sufficient for translation. As the discussion in
Chapter indicates, there are many cases where
problems seem to require deeper, more meaning oriented
representations, and enrichment of the kind of knowledge systems are
equipped with. In this chapter we will try to give a flavour of what
is involved in this.
It is useful to think of this knowledge as being of three kinds: (i)
linguistic knowledge which is independent of context , semantic
knowledge; (ii) linguistic knowledge which relates to the context
(e.g. of earlier utterances), sometimes called pragmatic
knowledge; and (iii) common sense , general, non-linguistic knowledge
about the real world , which we will call real world knowledge.
It should be stressed that the distinction between these different
kinds of knowledge is not always clear, and there are those who would
dispute whether the distinction is real. However, it is at least a
convenient subdivision of the field, and we will examine each sort of
knowledge in turn, in Sections
, , and .
Discussing these different kinds of knowledge will also allow us to
describe some more general translation problems.
Apart from giving an overview and flavour of what is involved, the point we would like to stress in this chapter is that though dealing with meaning in a general way poses many unsolved problems, and in general one should not expect to find much in the way of real world , pragmatic , or even semantic processing in current commercial MT systems, such processing it is not totally beyond the reach of current theory.