Information retrieval tasks concerns with leading the user to those documents that best match his/her information needs. We consider missing concepts and wrong terms in user query as the fundamental problem of information retrieval. We propose a multiagent system that assists information retrieval by mediating the user’s information needs and the semantic structure of the data domain. The multiagent system embeds both ontology and thesauri to traverse different cognitive spaces. During an interactive process, the user’s query is transformed and led to appropriate semantic constructs that enable effective retrieval. We consider the data domain of government official gazettes. A prototype system, Gaz-Guide, has been developed and experiments are conducted by recording system response to real users with practical questions. The initial results show encouraging sign of the utility and effectiveness of Gaz-Guide in articulating domain resources on thesauri and ontology and guiding users with interactive assistance.