Abstract:
The paper compares three methods for parsing of patients' chief complaints extracted from electronic medical cards. We propose two methods which are based on usage of an ontology: either as a method for correction of mistake made by a parser, or for constructing syntactical dependencies according to this ontology and a limited set of rules of syntactical governance. As a control test, we use existing natural text parsing libraries. The paper demonstrates that such a simple approach could achieve a high accuracy, which is comparable to modern parsers.
Keywords:parsing, shallow parsing, ontology, medical texts.