The concept processor should be able to extract a maximum of information from the user and store this information for later use. It should also be flexible enough to deal with a variety of concepts and with changes of concepts over time. The concept processor rests on the logic of sets. The program developed as a first step in implementing a concept processor applies Bayesian statistics to the word use in the documents to calculate the chance that a document is relevant. It also offers initial help to the user to decide which documents should be a-priori candidates for exemplars and counter-exemplars. Teachers have used the program to mark students. It has reduced the time normally required to mark as many as 200 assignments at a time. The similarity scores that the program uses have been useful in detecting fraudulent papers. A prototype has also been developed for classifying legal and other documents according to the topic. The system has potential usefulness for searching large databases relevant to areas such as criminal investigations or journalistic research.