The problem of ambiguity is pervasive in NLP, and is discussed extensively in the introductions to the subject such as those mentioned in the Further Reading section of Chapter .
Examples of lexical and structural mismatches are discussed in [Hutchins and Somers1992, Chapter 6,]. Problems of the venir-de/ have just sort are discussed extensively in the MT literature. A detailed discussion of the problem can be found in [Arnold et al. 1988], and in [Sadler1993]. On light verbs or support verbs, see [Danlos and Samvelian1992,Danlos1992].
Treatments of idioms in MT are given in [Arnold and Sadler1989], and [Schenk1986]. On collocations , see for example [Allerton1984], [Benson et al. 1986a], [Benson et al. 1986b] and [Hanks and Church1989]. The notion of lexical functions is due to Mel'cuk, see for example [Mel'cuk and Polguere1987,Mel'cuk and Zholkovsky1988].
A classic discussion of translation problems is [Vinay and Darbelnet1977]. This is concerned with translation problems as faced by humans, rather than machines, but it points out several of the problems mentioned here.
The discussion in this chapter touches on two issues of general linguistic and philosophical interest: to what extent human languages really do carve the world up differently, and whether there are some sentences in some languages which cannot be translated into other languages. As regards the first question, it seems as though there are some limits. For example, though languages carve the colour spectrum up rather differently, so there can be rather large differences between colour words in terms of their extensions, there seems to be a high level of agreement about `best instances'. That is, though the extension of English red, and Japanese akai is different, nevertheless, the colour which is regarded as the best instance of red by English speakers is the colour which is regarded as the best instance of akai by Japanese speakers. The seminal work on this topic is [Berlin and Kay1969], and see the title essay of [Pullum1991]. The second question is sometimes referred to as the question of effability, see [Katz1978,Keenan1978] for relevant discussion.