Preserving text categorization through translation

We treat natural language documents as we treat strongly-typed functional programming languages by introducing semantic categories as types. We use axioms to define primitive semantic categories and inference rules to capture the meaningful relationships among the primitive semantic categories. Primitive categories are nodes of a sense type decision tree. Axioms and inference rules are used to construct compound categories, validate category hypotheses, and eliminate ambiguities. The same categorization is obtained when this approach is applied to a text in a given natural language or to its translation into another language if a one-to-one mapping can be defined between the axioms and inference rules associated to the initial language and the axioms and inference rules associated to the other language.