论文部分内容阅读
Chinese microblog sentiment classification has become increasingly useful to reveal the emotional tendency on specific event or policy.But the current study is limited to emotional polarity classification,failing to extend the emotional strength judgment.And general emotion dictionary is unable to effectively implement the sentiment classification of specific domain topics.To solve these problems,in this paper,we adopt the lexical semantic similarity calculation model,which is based on HowNet,to extend the existing emotion dictionary,and then propose the calculation model of emotional tendency value based on weighted emotional tendency.Besides,we introduce view recognition into emotion classification to filter non-opinion sentences to improve the efficiency of data processing and classification quality.Experimental results indicate that our approach outperforms the previous method significantly.