Citation function classification based on ontologies and convolutional neural networks

Abstract

In recent years, there has been significant growth in the use of citation to improve the methods of evaluating the quality of publications. To determine the quality of the publications, traditional methods such as impact factor depend only on the citation count. Recently, citation functions or purposes have gained attention to evaluate the quality of these methods. Citation function classification is defined as a way to find out the reasons behind quoting previous literature. Several approaches for citation function classification have been proposed to classify citation functions in scholarly publication. However, these approaches do not consider the author’s characteristics such as author’s information, neither the publication level. Those characteristics can be useful in the process of citation function classification. In addition, previous studies mainly used classical machine learning techniques such as support vector machine and neural networks with a number of manually created features. The manual feature representation is time-consuming and error prone. To address these problems, we propose a citation function classification model by combining ontologies with convolutional neural networks (CNN). In our model, ontologies were used to represent the author’s characteristics and the citations semantically. Then, we have incorporated this representation into a CNN model to classify citations into six functions. We have conducted experiments using public dataset and showed that the proposed approach achieves good performance compared with the existing techniques in terms of accuracy.