基于序贯三支决策的多粒度情感分类方法

如何对评论数据进行正确的情感分类是情感分析中的重要研究内容。从粒计算和认知学角度,提出了一种基于序贯三支决策的多粒度中文评论情感分类方法。首先,基于评论数据集的特点,根据评论中情感信息量的多少,提出一种由粗到细的多粒度情感信息表示方法；然后,结合序贯三支决策的思想在不同粒度依据情感信息进行逐步计算,对边界域评论序贯地进行三支决策；最后,根据不同粒度的决策阈值和成本对评论做出最终的情感分类。对比实验结果表明,该方法在3个经典评论数据集上获得了更好的结果,具有更高的分类正确率和更强的鲁棒性。

Multi-granularity Sentiment Classification Method Based on Sequential Three-way Decisions

How to classify the review data correctly is important research content in sentiment analysis.From the perspective of granular computing and cognitive science,this paper proposed a multi-granularity sentiment classification method for Chinese reviews based on sequential three-way decisions.Firstly,based on the characteristics of review data,a coarse-to-fine multi-granularity sentiment information representation method is put forward according to the amounts of sentiment information existing in the review.Then,by combining the principle of sequential three-way decisions,the calculation is gradually executed in different sentiment information granularity and the sequenced three-way decision is carried out for the boundary reviews.Lastly,according to the decision thresholds and costs in different granularities,the final sentiment classification is provided for the review data.The experimental results show that the proposed method achieves better performance,and performes higher classification accuracy and stronger robustness on three classic datasets.