Our criterion of linguistically possible translation is defined as follows:
Machine translation finds its typical applications in non-literary texts, e.g. news bulletins, scientific articles, etc. It is obvious that in such texts, what is said about the world is the most important thing. For example, one would like the translation of a weather report or news bulletin to be true if and only if the original is true (even more so with airplane maintenance manuals).
However, the first criterion is too weak for practical applications. If for a given source text there is a target language equivalent , then a huge number of other target texts are equivalent too, like ` and ' where is a tautology. Similarly, all tautologies become translations of each other. A finer view of `meaning' is necessary. New developments (e.g. property theory ) may lead to more interesting criteria but it is not clear at this moment how to apply them in practice. Therefore we require also some preservation of syntax. The way in which meanings of expressions are composed from the meanings of their parts is preserved wherever possible. To take an example from Lewis [p. 182]lewis `Snow is white or it isn't' differs in `syntactic meaning' (though not in truth conditions) from `Grass is green or it isn't' because the embedded sentences differ in truth conditions. The same idea has been applied by Landsbergen .
This view of translation is extremely poor. For example, it does not take world knowledge into account. Therefore, a sentence like `the prime ministers discussed the situation of Iran in Moscow' is ambiguous in the MiMo2 system (even though the two meanings may sometimes lead to identical translations). Moreover, there are many other factors that could be taken into account in defining linguistically possible translation, e.g. preservation of style, (indirect) speech act, honorifics, etc. It is hoped (and expected) that an approach based on the poor view described can be useful as a basis for future richer views.
An important question of translation is whether there always is a meaning-preserving translation. It may be the case that there are meanings in one language that are not expressible at all in some other language (for some discussion cf. ). It may even be the case that one cannot know whether the meaning expressed in two languages is the same (cf. ). These are important questions, but the MiMo2 system is irrelevant to them. It is concerned only with the case where the same meaning can be expressed in both languages. Our question is `how to describe possible translations', not `is translation possible'.