An Adverse Drug Events Ontology Population from Text Using a Multi-class SVM Based Approach

摘要

In recent years, semantic web technologies and ontologies in particular, are being increasingly used in various e-Health systems and applications. However, issues related to automatically constructing, populating and enriching such ontologies are still outstanding. In this paper, we propose an automatic Adverse Drug Events (ADE) ontology population approach so called ADETermino. The proposed approach is based on Information Extraction methods and mainly aims to extract new concept instances and relationships from textual drug leaflets. It combines a Named-Entity Recognition (NER) system using lexical resources and a machine learning method using a multi-class Support Vector Machine (SVM) classifier for relations detection. Experiments were performed using 102 cardiac drug leaflets corresponding to 5706 input vectors. The results show the performance of our approach with an F-score of 89%.