摘要
介绍基于词性组合规则改进的中文句子极性判断方法,提出一种基于半监督学习的中文句子极性判断框架。在传统的完全基于情感词典方法的基础上,结合词性组合规则这一重要特征对中文句子进行极性判断。首先,分析中文句子中情感短语、情感词语的词性组合规则。然后,将情感短语、情感词语的词性组合规则用于中文句子极性判断。根据词性组合规则集抽取评测句子中的候选情感短语、情感词语;而后,计算句子的情感信息总量和句子的情感值,根据句子的情感信息总量将句子分为主观句、客观句,根据句子的情感值将主观句子分为积极情感句、消极情感句、中立情感句。实验结果证明,该方法在主客观分类上F值较高,可以达到77.4%;在主观句情感分类上,可达到的F值为62.5%。相比较于已有方法,基于词性组合规则改进的中文句子极性判断方法的F值有了明显的提高。
In this paper we introduce an improved Chinese sentences polarity judgment method which is based on the rules of part of speech（ POS） combination and present a semi-supervised learning based Chinese sentence polarity judgment framework. On the basis of traditional approach which is based solely on emotion dictionary,we combine an important feature named the rules of POS combination to make polarity judgment on Chinese sentences. First,we analyse the rules of POS combination of emotional phrases and emotional words.Then,we apply these rules in Chinese sentences polarity judgment. According to the POS combination rules set,we extract the candidate emotional phrase and emotional words in sentences to be evaluated and then calculate the total amount of emotional information in sentences and the emotional value of the sentences. Finally,according to total amount of emotional information in sentences,they will be distinguished as subjective sentences and objective sentences,and according to emotional value of the sentences,the subjective sentences will be divided into positive emotional sentences,negative emotional sentences and neutral emotional sentences. Experimental results show that the F value of this method in subjective and objective classifications is quite high,which can reach up to 77. 4%. In subjective sentence emotion classification,the F value of this method can achieve 62. 5%. Compared with existing methods,the F value of this Chinese sentences polarity judgment method based on the rules of POS combination has significant improvement.
出处
《计算机应用与软件》
CSCD
2015年第3期309-312,330,共5页
Computer Applications and Software
关键词
极性判断
词性组合规则
句子的情感信息总量
句子的情感值
Polarity judgment Rules of part of speech combination Total amount of emotional information in sentence Emotional value of sente
