Beliefs in Multi-Agent Worlds: a Terminological Logics Approach