Semantic analysis supported by inference in a functional model of language

We describe a model of semantic analysis (SemETAP) aiming at producing semantic structures of Russian sentences. The model is a new option of the ETAP-4 linguistic processor and reuses some of its linguistic knowledge. We believe that the more inferences we can draw from a text, the better we understand it. Therefore, the model tries to make all the inferences its knowledge allows us to make. The inferences are of two types: strict implications and plausible expectations (implicatures). We describe diff erent types of knowledge, both linguistic and extralinguistic, which the model has at its disposal, and show how this knowledge can be used for text understanding. In particular, we discuss some issues related to reference, which are diffi cult for automatic analysis and show how SemETAP helps cope with them.