A Natural Language based Knowledge Representation Method for Medical Diagnosis

Expert system can effectively reduce human error and improve the diagnostic quality. However, due to the medical domain knowledge is large and complex, effective knowledge representation or vocabularies standardization is important issues to ensure both shared understanding and interoperability between people and clinical decision support system (CDS). This paper uses a semantic model to convert natural language based web resources into machine understandable information. With an ontology, both natural language based user description and descriptive web knowledge can be mapped into same structure so that be able to calculate the similarity. Factors which will affect the system understanding ability are studied. Based on the results, an interface with reasoning function is provided to help the user refine his/her input incrementally. Experiments show that reasoner do improve the recognition ability of the CDS system.