Abstract

Online communication systems or bulletin board systems have beenwidely used before the age of the Internet began. Such communicationsystems accumulate human knowledge as messages submitted by the users.Since most data included in the systems are just plain texts, it isdifficult to reuse data just by employing simple pattern matchingtechniques. In this paper, we present a new online communicationsystem that contains highly-reusable semantically-annotated messages.Semantic annotation on messages is realized by user definable datastructures called message classes. The message class is defined usingXML (eXtensible Markup Language) and has functionality to generate anappropriate message input form. The users can write messages on theform, and then semantically-annotated (structured) messages areautomatically generated. Our online communication system can alsomanipulate the accumulated message data to search for some valuableinformation and to generate mined results by applying some operationsto whole or some part of the messages. For example, the system cangenerate a personalized ranking of a user's preferred restaurantsbased on reviews included in the messages. To flexibly manipulate themessage data, our system has a mechanism for filtering the messagesand for applying operations to the filtered messages. Such operationsare implemented by integrating user-customizable operators and Perlprograms. Therefore our system can not only accumulate humanknowledge as structured message data but also help the users todiscover useful information by filtering and mining the messages.