The article considers ontology as a set of relations (unary and binary) which are represented by specialized matrix like structures – C-systems. That allows us to consider tasks of inference on ontologies as constraint satisfaction problems. A method of a priori analysis and transformation of SPARQL queries patterns into a form which speeds up the subsequent execution of concrete user queries has been developed. The method is oriented on ontologies developed with using content ontology design patterns, that ensures the predictability of the structure of potential queries. The method is based on the combining of the methods of structural decomposition and the original methods for non-numerical constraints satisfaction.