基于融合词向量模型的特色文献分类  

Classification of Characteristic Literature Based on the Fusion Word Vector Model

作　　者：陈蓝 周杰 杨帆[2] CHEN Lan;ZHOU Jie;YANG Fan(Guizhou University Library,Guiyang 550025,China;Guizhou University of Finance and Economic Library,Guiyang 550025,China)

机构地区：[1]贵州大学图书馆,贵州贵阳550025 [2]贵州财经大学图书馆,贵州贵阳550025

出　　处：《现代信息科技》2025年第8期157-160,164,共5页Modern Information Technology

基　　金：贵州财经大学2022年度校级科研基金项目(2022KYYB14);2024年度贵州省高校人文社会科学研究项目(2024RW313)。

摘　　要：图书馆服务工作中,面对小数据量的地方特色文献,管理人员需要花费大量时间精力手动整理地方特色文献。为实现特色文献实现自动化预分类,文章提出了针对小数据量的文献自动化分类模型CSGB,以贵州省特色文献数据集为实验对象,通过GloVe及BERT进行预训练,对其产生的向量进行向量融合,通过TextCNN进行特征提取与表达,对不同数据规模的特色文献进行分类。实验结果显示,融合字向量的准确度相较于基准模型有至少4%的提升。In library service work,when facing local characteristic literature with a small data volume,library managers need to spend a great deal of time and effort manually organizing such local characteristic literature.In order to achieve automated pre-classification of characteristic literature,this paper proposes the CSGB model,which is an automated classification model for literature with a small data volume.Taking the characteristic literature dataset of Guizhou Province as the experimental object,the model conducts pre-training through GloVe and BERT,fuses the generated vectors,extracts and represents features through TextCNN,and classifies characteristic literature of different data scales.Experimental results indicate that the accuracy of the model with fused word vectors is at least 4%higher than that of the benchmark model.

关 键 词：地方特色文献 文本分类 文本向量化 

分 类 号：TP391.1[自动化与计算机技术—计算机应用技术]