Researchers

Islam Kheelan, Nada Ghneim and Ammar Joukhadar

Published in

Damascus University Journal for Engineering Sciences, First conference on Informatics, volume 38, issue 4, October 2022.

Abstract

The task of relation extraction is to find semantic relations between text entities and classify them according to the type of the relation. This paper aims to extract relations from Arabic texts using Distance Supervision, so that we initially collected the ArRe24k Arabic corpus automatically by adopting the Distance Supervision methodology, then we trained Piecewise Convolutional Neural Network (PCNN) model with a sentence-attention layer, to detect and classify the relations between the mentioned entities in the Arabic texts and overcome the errors of automatic labeling caused by the strong assumption of

Distance Supervision. The model on ArRe24k data showed acceptable results with accuracy (71.6%) using manual evaluation.

Keywords: Relation Extraction, ArabicText, Deep learning, Distance Supervision.

Link to full paper

http://journal.damascusuniversity.edu.sy/index.php/engj/article/view/6329/1564