Description: The search engine based on the multi-agent approach conception for e-learning systems is presented in the paper. The main functions of the multi-agent system of information retrieval of the learning resources are defined; the component architecture is proposed which is based on four types of agents: a crawler, an appraiser, an indexer and a coordinator. The agentcrawler transfers to the specified by agent-coordinator URL-address and extracts metadata, if they are available, additional data and transmits these data to the assessor-agent. The assessor uses a comparator based on the domain model and the descriptor search model, which provides an efficient mechanism for determining the prospects for further search from the evaluated page. The main task of the index agent is to automatically create metadata that describes the found document in accordance with the chosen model. The search for educational facilities is considered in three aspects: the search for archives (repositories), the search for educational content in one archive and the search for learning resources. Such tasks complement each other and together implement the search function in many archives, which is one of the important components of a collaborative e-learning system. Information retrieval for educational materials is an intellectual process and requires the development of a special ontology of search and subject ontologies in the field of study. The idea of the gradual narrowing of the search space is proposed. Narrowing of the search space is based on the usage of comparator identification. The comparator allows realizing the evaluation of the search result based on the equivalence predicate using the subject ontology. The obtained results can be used to implement the agent system of intellectual search for learning objects in e-learning systems.
Keywords: information retrieval, search engine, multi-agent approach, e-learning, descriptor search, comparator identification