Extraction of Relations between Entities from Texts by Learning Methods

摘要

The aim of this work is to automatically extract structured information from unstructured texts, permitting their fusion in an intelligence application. In Thales, we have a knowledge management system (Ideliance) that permits us to manage entities and relations between them, but at present the user must manually capture this information. To automate such an extraction, we propose the use of a learning algorithm that we have developed after the study of the existing information extraction methods. We present the Sem+ tool that implements the algorithm, and the evaluation of this tool carried out by us and by the Land Headquarter (S.T.A.T. unit).