基于BERT模型的指令集多标签分类研究  

Research on instruction set multi-label classification based on BERT model

作　　者：王淳睿 何先波[2] 易洋 WANG Chunrui;HE Xianbo;YI Yang(College of Electronics Information Engineering,China West Normal University,Nanchong Sichuan 637009,China;College of Computer,China West Normal University,Nanchong Sichuan 637009,China)

机构地区：[1]西华师范大学电子信息工程学院,四川南充637009 [2]西华师范大学计算机学院,四川南充637009

出　　处：《智能计算机与应用》2022年第10期75-78,85,共5页Intelligent Computer and Applications

基　　金：西华师范大学英才科研基金项目(17YC149)。

摘　　要：指令分类可有效识别出指令的类别并提取出指令中的关键信息,有助于编译器开发人员快速构建编译器后端指令相关的代码,降低时间成本。针对传统文本分类模型的不足,研究提出基于BERT预训练模型的指令描述文本分类方法。经过数据清洗、预处理,在BERT预训练模型基础上,研究构建了文本多标签分类模型,实验结果表明,本文提出的方法对指令文本有较好的分类效果。The instruction classification can effectively identify the category of the instruction and extract the key information in the instruction, which helps the compiler developer to quickly construct the code related to the back-end instruction of the compiler and reduce the time cost. Aiming at the shortcomings of traditional text classification models, this paper proposes an instruction description text classification method based on the BERT pre-training model. After data cleaning and preprocessing, based on the BERT pre-training model, a text multi-label classification model is researched and constructed. The experimental results show that the method proposed in this paper has a good classification effect on instruction text classification.

关 键 词：深度学习 BERT 多标签分类 指令集 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]