Online social networks are being used by more and more people. While they enable rich communication, they cause privacy problems. One problem that has not been previously researched is the revelation of private information on blogs and comment parts of online social networks. This paper describes a new disclosure control system for natural language information that automatically checks blogs and comments uploaded onto online social networks, detects words that might reveal private information, and modifies the text by replacing the problematic words with more abstract words so that it no longer reveals the private information. The granularity of the proposed disclosure control is not the text but each word in it. Consequently, it is not tiresome for the user and balances the protection of privacy with the enjoyment of communication.