The system of semantic analysis of response texts in natural language

This article describes a two-level model of students answers evaluation, which serves as a basis for the creation of the system of semantic analysis of response texts in natural language in a dialog context. Two important methodological principles that make this system efficient are explained: «determinism of context» and «expectedness of the meaning of the answer». This article outlines the architecture of the system and six basic principles of its implementation: marking out of semantic units system, semantic classification of question-answer texts based on standard relations, development of individual conceptual grammars of semantic classes, segmentation of question-answer texts, relevance of knowledge representation (of the response model), openness of the system. The performance of the system is shown on the example of the analysis of an answer of the «Function» class.