This paper focuses on lexical semantics, polysemy and disambiguation in the perspective of medical language understanding. Previous results obtained in the domain of transplantation and organ failure with the semantic analysis of medical terms prompt us to check the capabilities of our prototype and its underlying semiotic model to deal with ambiguities and polysemy. We show how experimentations around linguistic observations lead us to introduce a void type to respect lexical ambiguities when the co-text is not sufficient for disambiguation. Then we demonstrate how the "creative use of words' (i.e. new senses in novel context) imposes to dynamically allocate types and roles according the co-text, using a type contraction operation.