For many language resources it is necessary to identify more than one language. In some cases, such as MT systems and bilingual lexicons, there is an added complication, namely directionality. Someone looking for such a resource will usually want to specify source and/or target languages. For example, in searching for a Korean-to-English resource, the user would not usually be interested in discovering English-to-Korean resources. Thus, there is an a priori need for resources to be described using ordered pairs of languages.
OLAC provides some mechanisms which can be applied to these cases. First, OLAC metadata elements are repeatable, e.g. a system which can process multiple languages will be described using multiple Subject.language elements, one per language. Second, OLAC incorporates directionality in its distinction between Language and Subject.language. Are these simple mechanisms adequate for the convenient and accurate description and discovery of multilingual resources? In this section we enumerate the main types of multilingual resource and show how OLAC metadata can be used to classify them. We report some problems and discuss possible solutions.