Classification of diagnoses that are described in natural language

Byline: Berry (L.M.) de Bruijn, Arie Hasman, Jan Willem Arends A method is presented to classify pathology reports that were composed in natural language. This way, the pathologist need not formulate a formal classification, but can adopt automatically suggested classifications. This paper describes the underlying theoretical aspects of the nearest neighbour method that was used, gives considerations for practical use, and presents the results of two studies on the method.