Since introduction in the 1970s concept maps have become more and more widespread in the field of teaching and learning. A concept map is a graphical tool for knowledge representation that facilitates automated knowledge assessment. To assess learner’s created concept map in an automated way, computer system must be able to evaluate conceptual structures different from expert’s concept map. Adaptive intelligent knowledge assessment system (IKAS) that has been developed in Riga Technical University since 2005 uses graph patterns to expand expert’s concept map. Graph patterns allow to infer hidden relationships from expert’s map based on linking phrases and to score them if they appear in a learner’s concept map. In this paper the extension of the set of linking phrases is described, appropriate graph patterns and patterns for scoring inverse relations are introduced to increase the automation of assessment.