Text mining: riconoscimento di elementi diagnostici in lettere di dimissione ospedaliera
Wednesday 6th April 2011
Text mining is a quite recent research area in computer science whose aim is to automatically provide information from electronic textual resources. It combines statistical, linguistic and machine learning techniques to mine unstructured data as text documents. In this work, an overview of text mining techniques is given, together with a general presentation of possible applications. Moreover, an information extraction system is built to mine diagnostic medical reports. The process consists of recognizing specific elements that indicate a myocardial infarction diagnosis, using a machine learning approach based on SVM algorithms. Results are presented, focusing on the influence of a priori linguistic knowledge used to define the machine learning system.