Learning by Constraint Relaxation
We present a soft computing approach to text processing and propose a constraint theoretic approach for machine learning called learning by constraint relaxation (LCR). Words do not convey exact meanings but act as soft constraints over possible meanings. LCR is based on this principle: a failure to recognize is an opportunity to learn. A rule-making language, RML, has been implemented to facilitate problem representation and experimentation with constraint relaxation problems. LCR has been used to extract knowledge from texts containing medical descriptions and informatics codes.
KeywordsConstraint Satisfaction Constraint Satisfaction Problem Soft Constraint Text Extraction Constraint Relaxation
Unable to display preview. Download preview PDF.
- 1.Barták, R.: Constraint Programming: In Pursuit of the Holy Grail. In: Proceedings of Week of Doctoral Students. Charles University, Prague (1999).Google Scholar
- 2.Blum, A., Mitchell, T.: Combining labeled and unlabeled data with co-training. In: Proceedings of Conference on Learning Theory. (1998).Google Scholar
- 5.Freuder, E.: Partial constraint satisfaction. In: Proceedings of the International Joint Conference on Artificial Intelligence. (1989) 278–283.Google Scholar
- 7.Winston, P.H., Horn, Horn, B.K.P.: Lisp, Third Edition. Addison-Wesley. (1988).Google Scholar