Building a model of disease symptoms using text processing and learning from examples

The paper describes a methodology of building a semantic model of disease symptoms. The fundamental techniques used for creating the model are text analysis and learning from examples. The text analyser is used for extracting a set of symptom descriptions. The descriptions are a foundation for delivering a user interface, necessary for collecting patient cases. Given the cases a semantic model is built, which is achieved through clusterisation and statistical analysis of cases. The approach to creating the model eliminates the need of direct model manipulation, because the meaning is retrieved from association to diseases instead of purely linquistic interpretation of symptom descriptions. Detection of synonyms is also completely automatized.