Multi-task learning model based on recurrent convolutional neural networks for citation sentiment and purpose classification

Automated citation analysis is a method of identifying sentiment and purpose of citations in the citing works. Most of the existing approaches use machine learning techniques to boost the performance of citation sentiment classification (CSC) and citation purpose classification (CPC), which are the main tasks of automated citation analysis. However, such approaches address CPC and CSC by learning them separately, which often suffer from inadequate training data and time-consuming for feature engineering. To alleviate these problems, we propose a multitask learning model based on convolutional and recurrent neural networks. The proposed model benefits from jointly learning CSC and CPC by modeling the citation context with task-specific information and shared layers for citation sentiment and purpose classification. The network architecture of the proposed model is useful to represent the citation context and extracts the features automatically. We conduct experiments on two public datasets to evaluate the performance of the proposed model using standard metrics such as precision, recall, and F-score. The results of CSC and CPC tasks show improvements relative to classical machine learning algorithms such as SVM and NB as well as single-task deep learning models.