Preliminary Recommendations

The concept of common category

A debate among the MULTEXT partners generated some possible definitions of common category on which it is worth reflecting:

• The same meaning of a category in different languages implies that it enters into the same combinations (distributional definition of common categories).
• A category is common if it yields isomorphic partitions of word-lists or lexicons under translation (modulo exceptions). A category system is common to languages L and L if members of class C in L translate as members of C in L (translational definition of common categories);
• A common category among languages is a category which conveys the same linguistic information (i.e. stands for the same linguistic phenomenon) in all the languages in which it is used;
• A word class is common across a set of languages if it reflects a linguistic category or phenomenon which is either morphologically or lexically expressed in at least two of these languages, even though there may not be a one-to-one lexical relationship for this in translation.

We could define common categories as

Common categories: Categories whose members satisfy the same criteria and tests.

This crucially implies a clear definition of criteria for the recognition of their members (see the guidelines produced in the validation phase). In the absence of such explicit criteria, we can empirically recognise common categories which are relevant in the morphosyntactic descriptions of a number of European languages. These common categories are usable, are actually used in the largest lexicons and corpora, and have in general the same meaning in the different languages, even though the property of commonality holds more for open classes and poses more problems for function words or closed classes.

Keeping in mind this simple equation of common categories and actually used categories, their adequacy in terms of user requirements can also be achieved: this is empirically obtained through the bottom-up process of looking at the largest and most used lexicons and tagging schemes (as was done here). There have been many different types of users of these lexicons and annotation schemes.

In fact, with the above caveat in mind, it was found from the analysis of the schemes that many of the commonalities here proposed are relevant for many languages. Different schools and traditions of languages can agree on such a simple set of features.