An approach in health relation extraction

Abstract

Extracting relations among medical concepts is very important in the medical field. The relations denote the events or the possible relations between the concepts. Information about these relations provides users with a full view of medical problems. This helps physicians and health-care practitioners make effective decisions and minimize errors in the treatment process. This paper collects methods for relations extraction in health texts and presents an approach on one type of specific relation (i.e. template filling). The approach combines methods including rule-based and machine learning-based. The rule-based method uses the relation of semantic dependencies among the concepts to extract the rule set. The machine learning-based method uses the SVM (Support Vector Machine) algorithm and a feature set proposed. The results of the approach were estimated on an accuracy of 0.849.