Citation classification using multitask convolutional neural network model

In the recent years, there has been an increased availability of scientific publications across the world connected through citations. To help analyze this huge amount of information, citation classification has been introduced to identify the opinions and purposes of the authors for citing earlier works. Existing approaches utilize machine learning techniques and report promising results in identifying the sentiment and purpose of the citations. However, most of the previous approaches tackle the citation sentiments and purposes classification in isolation. Moreover, they suffer from limited training data and time-consuming feature engineering process. In this paper, we address these issues by building a multitask learning model based on convolutional neural network. The proposed model jointly learns the citation sentiment classification (primary task) with the citation purpose classification as a related task to boost the classification performance. Experimental results on two public datasets show that our model outperforms the previous baseline methods and prove the effectiveness of multitask learning technique.