Abstract: Abstract To cope with the mismatch problem of various expressions of the same concept in different information resources,an approach of structural description adaptive to the task circumstance as semantic indexes to the information body content is presented to meet the requirement of expanded query. Two mechanisms to realize expanded query, namely knowledge networking and compatible matching are introduced, correlating to the top-down access and bottom-up access respectively. Moreover， query rewriting template (QRT) is exploited to search the useful knowledge related to the present task. QRT produces a large number of sub-queries based on the original query and ontology,at the same time passes the weight values to sub-queries which score their degree of relativity to the original query. Top-down approach or knowledge networking mechanism retrieves the related knowledge items by organization and task ontology, while bottom-up approach searches the similar task in task circumstance and retrieves the knowledge items containing the task description. In both cases query rewriting template is applied to realize the ontology-based knowledge retrieval. Experiments show that the proposed approach improves the retrieval efficiency and accuracy of knowledge management system.