摘要
文本分析领域内传统的分类模型大多基于情感词典和概率统计模型,没有同文本语义建立联系。隐马尔可夫模型(Hidden Markov Model,HMM)能够体现不同类别文档的语法和语义特征,在文本处理领域表现优异,因此逐渐取代传统分类模型,成为新的研究方向。该文采用不同情感类别的训练文本来构建HMM分类器,并通过自适应增强(Adaptive Boosting,AdaBoost)算法进一步提高HMM分类器的性能。实验结果表明,使用两者结合形成的Ada-HMM模型对评价类文本进行情感分类,分类准确率达到93.3%,优于传统情感分类模型。
The traditional classification models in the text analysis field are mostly based on sentiment dictionaries and probability statistical models,and have no connection with text semantics.Hidden Markov Model(HMM)can reflect the grammatical and semantic characteristics of different types of documents,and has excellent performance in the field of text processing.Therefore,it gradually replaces the traditional classification model and becomes a new research direction.This paper uses training texts of different emotion categories to construct HMM classifiers,and further improves the performance of HMM classifiers through adaptive boosting(AdaBoost)algorithm.Experimental results show that the Ada-HMM model formed by the combination of the two is used to classify the evaluation text with a classification accuracy of 93.3%,which is better than the traditional sentiment classification model.
作者
李秋伶
郑静
LI Qiuling;ZHENG Jing(School of Economics,Hangzhou Dianzi University,Hangzhou Zhejiang 310018,China)
出处
《杭州电子科技大学学报（自然科学版）》
2020年第6期50-55,共6页
Journal of Hangzhou Dianzi University：Natural Sciences
基金
国家社会科学基金资助项目(18BTJ026)。
