通过对微博文本的特性分析,提取了中文微博情感分析的关键问题:如何识别微博新词并理解其情感含义?如何利用附加信息辅助文本情感分析?如何结合语言特性构造情感计算方法?针对第一个问题,利用统计信息和点间互信息对新词进行挖掘和情感识别,在40万条新浪微博数据中构建了新情感词词典,用于对已有情感词资源的扩充。对于后两个问题,提出了基于词典和规则集的中文微博情感分析方法。根据微博特性,在不同的语言层次上定义了规则,结合情感词典对微博文本进行了从词语到句子的多粒度情感计算,并以表情符号作为情感计算的辅助元素。通过对采集到的原创微博数据集进行实验,验证了该方法的有效性。
This paper identifies the key points of Chinese micro-blog sentiment analysis:How to recognize the new words in micro-blog and comprehend their sentimental implication automatically? How to take advantage of additional informa-tion to assist text analysis? And how to construct a structure method based on micro-blog’s semantic characteristic? To solve the first problem, this paper utilizes a method using statistical information and point-wise mutual information to imple-ment new words mining and sentiment comprehension, and builds up a sentimental lexicon of new words from 40 million Sina micro-blogs to expand the existing resources. The sentiment analysis method introduced in the paper, based on rule set and lexicon, can cope with the rest questions. With rules defined on different semantic levels and optimized lexicon, micro-blog’s sentiment would be calculated on multi-granularity from words to sentences, adding emoticon information as auxiliary element of the total calculation. This method is conducted on original micro blogs of collected data base, and the result demonstrates its effectiveness.