领域同义特征词的统计规律及其在情感分析上的应用研究

Statistical law of domain synonymous feature-words and its application research on sentiment analysis

为了研究同义特征词的规律及其对情感分析的影响,基于大规模语料库的统计学习方法,发现如果将每一个特征词的所有同义特征词组成特征词组,词组中的同义特征词按照自身在语料库中出现的频率从大到小排列,那么同义特征词在词组中出现的频率与该词在词组中的名次存在幂函数关系:y=β0xβ1。将特征代表词的同义特征词所对应的情感词加入到该特征代表词的情感词集中,可以有效提高特征代表词的情感分析的准确性。

In order to study the law of domain synonymous feature-words and its influence on sentiment analysis,this paper based on statistical learning method of the large-scale corpus and found that if all the synonymous feature-words of each feature-word formed phrases where the synonymous feature-words of each phrase rank according to their frequency in the corpus from high to low,there existed a power function among the frequencies of the synonymous feature-word and the ranking of the synon-ymous feature-word in the phrase：y=β0 x^β1 .Moreover,this paper show that adding the emotional words of the synonymous fea-ture-words of the rep feature-word to the emotional word-set of the rep feature-word can effectively improve the accuracy of the sentiment analysis of the rep feature-word.