Guessing the meaning of neoclassical compounds within LG : the case of pathology nouns

This paper investigates the semantic aspects of the formation of a special kind of nouns, and proposes a way of guessing their meaning by making lexical semantics principles interact with word formation theories. The crucial issue is that the nouns under study, which constitute a subset of the so-called neoclassical compounds, are frequently and productively found only in biomedical domain corpora. So, they have never been taken into consideration by morphological theories for general language, because of their only recent availability within (often) online resources. Consequently, computing their structure and meaning often requires the definition of specific rules. The paper focuses on the example of compound nouns which can be linearly represented by PREF YX I, where PREF represents a quantifying prefix, X is the compound head, Y its extension, and –I, a suffix indicating pathology and standing for eg –ie (in French and German) or ia (in English, Spanish and Italian). First, we will see how words satisfying this surface structure obey peculiar word formation rules; second, the paper shows in which way their meaning is predictible, since it regularly varies according to interactions between PREF and the relation that Y holds with X. Finally, the paper proposes three language-independent rules sufficient to predict the meaning of all PREF (Y)X I pathology nouns, by adapting a device developed within the Generative Lexicon Theory (GLT)