Semantic recognition tools for ontology learning Intellectual algorithms can't be used without effective ontology learning procedures. Especially it concerns to text mining and knowledge discovery tasks. Although it isn't obvious, but an intellectual recognition expects motivation which estimation unavoidably involves planning processes with a target state expected utility estimation. Whole such structures and mechanisms are included into ontology of an intellectual agent and based on it its knowledge base. Ontology learning today stays unsolved problem because of absence of effective knowledge recognition algorithms, lack of right, planning-oriented ontology architecture and wide variety of domain specific demands to knowledge content which must be satisfied in-situ by learning procedures. Such recognition algorithms must be hierarchical [ ] – to recognize concepts and semantic relations on a lower level, predicates on a second level, rules – on a third level, states and tasks – on a forth, and at the end, task solving strategies at a highest fifth level of recognition (Fig.1). Fig. 1. Five levels of a knowledge recognition Such hierarchy of recognition anticipates complex system of means closely connected with (integrated into) knowledge base of recognizing agent. Each level of recognition must have its own set of evidences and decision making criteria, included into relevant knowledge base subsystem, but they are interconnected with aim to avoid natural language ambiguity problem, discover information inconsistency and estimate its importance and credibility. On a lowest recognition levels must be used specific knowledge concerned to natural language understanding, especially, vocabularies and syntactic parsers which produce data for recognition concepts and semantic relations. There are different tools that fulfill that role and could be included into whole ontology learning system as a part of semantic recognition tool. But parsed data must be deeply processed to extract known concepts and relations or learn new one by the system. Therefore there is a task to build a low level ontology learning subsystem with the task to adopt concepts and semantic relations from natural language texts ready to form appropriate predicates in terms of description logic using additional information from knowledge base.