Abstract:
Case base reasoning is an artificial intelligence (AI) paradigm that has proven to be a relevant technique in various fields of reasoning especially medical diagnosis. The technique provides solutions/answers to new problem through comparison of repository of finite set of known solved cases. Cases in the repository/case-store that share similar features with the new problem are retrieved and compared using some forms of distance measurement and any case with very low distance score will be adopted for consideration is solving the new case. Case base reasoning is however challenged by approaches for an exhaustive extraction and formalization of features in different cases. Failure to efficiently extract and formalize features of cases for reasoning produce a wave or repel effect on the distance measurement
and distance score, thereby yielding and inaccurate selection of solution-based case. This paper therefore presents a novel approach to feature extraction and formalization from text-based diagnosed cases of breast cancer. The paper uses the concept of natural language processing for feature extraction and formalization in description logic (DL) which translates into ontology learning (automated ontology building). Furthermore, we proposed a shift from the use of standard distance metrics (e.g. Euclidean distance) in case-features
similarity measurement to the use of ontology operation named ontology matching. As a result, we produced a model for efficient case base reasoning that is able to capture successful and failed diagnosis of breast cancer in solving new cases. The model represents a new case and case-base knowledge bases in OWL ontology format, and exploits the benefits knowledge integration by ontology through the integration of cases for larger case-based KB.
Keywords: Case based reasoning, description logic (DL), natural language processing (NLP).
