文本情感分析作为自然语言处理领域中的一个重要分支,在现实生活中具有重要的应用价值.传统深度学习模型在情感分析中主要根据概率值大小进行硬分类,忽略了极性不明显数据的影响,导致阈值边缘对象的分类准确率欠佳.为了解决这一问题,基于 CNN(Convolutional Neural Networks)和 BiLSTM(Bi-directional Long Short-Term Memory)模型,并引入序贯三支决策(Sequential Three-way Decisions,S3WD)的思想,提出了基于BiLSTM和CNN的序贯三支情感分类模型(BiLCNN-S3WD),该模型能更好地从多个粒度对极性不明显数据进行处理.通过在online_shopping_10_cat和微博数据集上进行多组实验与对比分析,验证了所提模型的有效性.实验结果表明,与七个基线模型相比,BiLCNN-S3WD在三个数据集上的每个评价标准都取得了更佳的效果.
Research on sequential three-way sentiment classification model based on BiLSTM and CNN
Text sentiment analysis is an important branch of natural language processing with significant application value.Traditional deep learning models in sentiment analysis mainly perform hard classification based on the size of probability values,neglecting the impact of the data with inconspicuous polarity and resulting in poor accuracy of the classification for threshold edge objects.Based on CNN(Convolutional Neural Networks)and BiLSTM(Bi-directional Long Short-Term Memory),we propose the BiLCNN-S3WD based on BiLSTM and CNN,by introducing the idea of S3WD(Sequential Three-way Decisions),which better processes the data with inconspicuous polarity from multiple granularities.The model's effectiveness is verified through multiple sets of experiments and comparative analyses on the online_shopping_10_cat and Weibo datasets.According to the experimental results,BiLCNN-S3WD achieves better results in each evaluation criterion on the three datasets compared with the seven baseline models.
国家科技期刊平台
NSTL
万方数据
