Abstract
In this paperwe explain how to derive learnability of some classes of categorial languages as a consequence of the a property that have unifiable infinite sets of types, namely unification compactness. At first we present an alternate (with respect to [6]) proof of learnability of rigid languages, that introduces our method. Then we prove learnability of the class of semi-rigid languages, which lies in between (learnable) class of rigid languages and (not learnable) class of optimal (as defined by Kanazawa in [6]) languages.