Neural machine translation for in-text citation classification
Text
ClassificationBackground: Citations have long been used to characterize the state of a scientific field and to identify influential works. However, writers use citations for different purposes, and this varied purpose influences uptake by future scholars. We conclude these purposes to five citation function BACKGROUND, MOTIVATION, USES, COMPARISON OR CONTRAST,FUTURE. BACKGROUND means cited paper P provides relevant information for this domain. MOTIVATION means P illustrates need for data, goals, methods, etc. USES means uses data, methods, etc., from P . EXTENSION means Extends P’s data, methods, etc. COMPARISON OR CONTRAST means Expresses similarity/differences to P. FUTURE means P is a potential avenue for future work. Labeling process: 1)Download literature paper and Extract citations; 2)While extracting We pre-annotation each citation and skip any instances whose context was corrupted .3)Conduct secondary annotation and inspection after pre-annotation. Train and test:Ultimately, we extracted and annotated 3,516 citations from 167 articles.The proportions of the five sections, BACKGROUND, MOTIVATION, USES, COMPARISON OR CONTRAST, and FUTURE, are approximately 10:3:3:1:1:1. In this dataset Then we conducted training and testing on this dataset using data augmentation and the BERT model. The test results indicated that the performance of the new model surpassed that of classic methods such as CNN, LSTM, and BiLSTM.