Fine-grained citation count prediction via a transformer-based model with among-attention mechanism
Shengzhi Huang,Yong Huang,Yi Bu,Wei Lu,Jiajia Qian,Dan Wang
DOI: https://doi.org/10.1016/j.ipm.2021.102799
2022-03-01
Abstract:Previous studies have confirmed that citation mention and location reveal different contributions of the cited articles, and that both are significant in scientific research evaluation. However, traditional citation count prediction only focuses on predicting citation frequency. In this paper, we propose a novel fine-grained citation count prediction task (FGCCP), which aims to predict in-text citation count from each structural function of a paper separately. Specifically, we treated this task as a “sequence to sequence” issue and a multi-task learning job, in which both the inputs and the outputs are based on the sequence pattern of citations from different structural functions. To fulfill FGCCP, we proposed a transformer-based model (i.e. MTAT) in which a novel among-attention mechanism is employed. Based on an empirical study of full-text documents from PubMed Central Open Access Subset, our model achieves satisfactory prediction accuracy, and surpasses common machine learning and deep learning models on FGCCP. Moreover, we also discuss the potential role of the among-attention mechanism and the reason why our proposed model outperforms state-of-the-art strategies. FGCCP may provide more detailed decision-making evidence and evaluation basis for researchers in scientific research evaluation. In addition, MTAT is a general model which can be easily deployed in other multi-task learning jobs.
computer science, information systems,information science & library science