Feed-Forward Network Model for Multi-Document Relation Classification

Abstract

Using feed-forward artificial neural network to classify multi-document relation is the subject of this paper. Sentences across topically related documents can often be linked by means of relations that exist between them. In this study, we aim to identify four types of relations, namely, Identity, Subsumption, Description and Overlap. We propose to use neural network learning model for the task of classification; multi-class classification, in this case. The performance of our proposed approach was measured using Precision, Recall and F-measure. The experimental findings demonstrate that better results can be obtained by using the proposed approach when compared with the widely used SVM classifier.