开放同行评审中自动评审分类方法研究OA北大核心CHSSCDCSSCICSTPCD

[Purpose/Significance]Scientific papers play a crucial role in the transmission and exchange of knowledge within academia.The evaluation of scientific paper reviews serves as an indicator of the knowledge value contained in these papers.Efficient and accurate prediction of scientific paper review classifications can enable swift assessment of their worth,thereby expediting the dissemination process for valuable knowledge.[Method/Process]This study delved into an automat-ic review classification method within open peer review systems.By harnessing semantic information extracted from scientific papers and expert ratings obtained during open peer reviews,the study constructed text representations and classification models.Traditional machine-learning approaches and deep-learning techniques were employed to generate automatic review classification results.[Result/Conclusion]Experimental findings demonstrate that integrating semantic information with rating data led to more effective review classification models compared to relying solely on mean ratings for judgment purpo-ses.Among the various models tested,the quality review classification model based on SCIBERT with the input of rating+mean achieved the highest accuracy at 90.17%.The proposed automatic review classification method demonstrated usability and high accuracy,offering valuable assistance to journal editors in swiftly screening potential scientific papers and contrib-uting to intelligent advancements in the field of scientific paper reviewing.