目前，常用的情感词典为通用情感词典。对于这些情感词典会存在如下缺陷，如词的情感区分度不够，对于不同数据集适应性欠佳。针对以上问题，基于情感词的情感确定性，提出一种新的权值计算方法。即情感确定性越大，其权值也越大，词的情感倾向也更明确，反之亦然。这样，使得字典内的情感词形成不同的区分度，这也与现实相符合。基于情感词权值，使用相应评分方法，运用到文本情感极性分析中。通过实验结果表明，这种方法相比只使用基础情感词典进行情感极性判断，准确度提升大约8．9%。这也说明了这种权值计算方法是有效的，结果符合预期。%Currently, the common emotion dictionaries are used frequently. But, There is some defects in these emotion dictionaries, such as not enough discrimination of the term, and poor adaptability to data set. To solve these problems, this article proposed a new method to calculate the weight of emotional terms with emotion certainty. The greater the emotion certainty of the term is, the bigger the weight is, and the more explicit the sentiment is. Using emotional weight will make term in emotion dictionary form the different degree of differentiation, which is also consistent with reality. And based on the weight of the emotional term, the appropriate scoring method was used to the analysis of text sentiment polarity compared to the method only using basic emotion dictionary. The experimental results show that this method would upgrade about 8. 9% on accuracy for emotional polarity judgment, which illustrates this weight calculation method is effective, and consistent with expectation.
展开▼
