A Multilabel Classification Method for Chinese Book Subjects Based on the Knowledge Fusion Model ERNIE-RCNN

At present, my country’s education modernization and university discipline construction plans have been carried out in major universities across the country. Classifying books according to the subject catalog can play an important role in promoting the discipline construction of universities, and it is also an important link in the construction of smart libraries. The traditional book classification work mainly relies on manual classification of books according to the Chinese Library Classification and some machine learning book classification algorithms, but it relies on domain experts to formulate rules, making migration difficult. This paper proposes a method of using the knowledge fusion model ERNIE-RCNN for multilabel classification of Chinese book subjects and constructs a Chinese book data set containing 82 subject labels to solve the problem that the current book subject label data sets are fewer and the classification effect is not good. Question. In the experimental results, the micro-F1 of the ERNIE-RCNN model is 0.8091, which has achieved a good classification effect. This study is helpful to realize the classification and organization of books according to the subject catalog and has a positive effect on the high-quality subject service provided by university libraries.