Automated Event Extraction Model for Linked Portuguese Documents

Abstract

In recent times, Machine Learning is booming and researchers are applying it to the most conceivable cases such as the area of linked documents. This article presents a process of automatic event extraction from Portuguese linked document whose accuracy (95.00%) was calculated by manual verification. With the help of an ontological structure, extracted events are mapped as a knowledge graph that represents the named entities and the events associated with each document. Such graphs are accessible through SPARQL queries. This way, the information existing in the linked documents can be easily accessed by resorting to a question-answering approach.