科技创新与应用2025，Vol.15Issue(12)：48-51,4.DOI:10.19981/j.CN23-1581/G3.2025.12.011
一种基于BERT的领域情感词典构建方法
摘要
Abstract
Aiming at the problem that the domain sentiment dictionary construction method based on seed word expansion has low ability to extract word context semantic information,which leads to poor domainness of the constructed domain sentiment dictionary,a BERT-based domain sentiment dictionary construction method(BERT-SentiL)is proposed.This method regards the construction of a domain sentiment dictionary as an emotion classification task of words or phrases,uses the BERT model to extract the semantic feature representations of candidate sentiment words,and then inputs the obtained semantic feature representations into Softtamx to obtain the confidence(probability)of the category(positive and negative)corresponding to each candidate word.Finally,the candidate words are emotionally classified according to the confidence to build a domain sentiment dictionary.Experimental results show that the domain sentiment dictionary constructed using the BERT-SentiL method has good domain expressiveness and emotion recognition effectiveness.关键词
情感词典构建/BERT/词向量/情感分析/语义特征Key words
construction of sentiment dictionary/BERT/word vector/sentiment analysis/semantic features分类
信息技术与安全科学引用本文复制引用
谢云高..一种基于BERT的领域情感词典构建方法[J].科技创新与应用,2025,15(12):48-51,4.基金项目
湖南省教育厅科学研究项目(23C0805) （23C0805）
湖南铁道职业技术学院校级一般课题(KJ202201) （KJ202201）
