A compound is a combination of two or more words which functions as a single word. In English, the most common type of compound is probably a compound made up of two nouns (noun-noun compounds), such as those in the dictionary entry for button:
In Spanish , for example, other types of compound s are equally important, including adjective-adjective compounds:
Orthographically , different languages follow different conventions for compounds . For example, in German compounds are generally written as one word, but in English some are written as one word (as buttonhole and buttonhook above), some as hyphenated words (e.g. small-scale) and some as juxtaposed words (e.g. button mushroom).
As with derivations, it is possible to describe the range of possible compounds by means of a word grammar, and as with derivations the possibility that one might be able to translate compounds by translating the component parts is very attractive --- especially since it is in principle not possible to list all English compounds, because compounding can give rise to words that are arbitrarily long. To see this, consider that one can form, in addition to film society:
Unfortunately, though there are cases where decomposing a compound and translating its parts gives correct results (e.g. the German compound Wassersportverein translates as water sport club), the problems of interpretation and translation are even worse for compounds than for derivation s. Apart from the fact that some compounds have completely idiosyncratic interpretations (e.g. a redhead is a person with ginger coloured hair), there are problems of ambiguity . For example, student film society could have either of the structures indicated, with different interpretations (the first might denote a society for student films, the second a film society for students):
A different type of ambiguity can be illustrated by giving an example: satellite observation may on one occasion of use mean observation by satellite, while on another occasion of use it might mean observation of satellites. Most of the time humans are able to rely on either our world knowledge or on the context to unravel a compound's meaning. Moreover, it is frequently important for translation purposes to work out the exact relation expressed by a compound. In Romance languages, for example, this relation may be explicitly realised by a preposition. For example, research infrastructure in Spanish translates as infraestructura para la investigación (literally, `infrastructure for research'). Nor can we happily assume that an ambiguity in one language will be preserved in another. Thus satellite observation has two possible translations in Spanish , depending on its meaning: observación por satelite (`observation by satellite') and observación de satelites (`observation of satellites').
A further problem with compounds is that a wide variety of relations are possible between the elements of a compound. Thus buttonhole is a hole for buttons, but button mushroom is a mushroom that resembles a button. It is not clear how to capture these relations.
Thus, as with derivation s, a really general approach to the treatment of compounds remains a research goal for MT.