一种网络舆情的定量分析方法*

分析公众通过个人博客、BBS和论坛表达出来的对于某一话题的态度、情感和观点倾向，对于党和政府及时掌握社情民意具有十分重要的意义。现有的舆情分析方法根据文本中表达的观点倾向将其分为肯定和否定两类，不能进一步给出肯定和否定的程度。提出了一种网络舆情的定量分析方法，该方法将对于某话题的文本集合作为输入，输出一个实数表示文本中所表达的观点倾向程度。实验表明，该方法的准确性能达到人工判断水平。

Quantitative method for analyzing public opinions on Internet

Analyzing the opinions expressed by the public in Web logs, forum, and BBS is very important for policy-makers. The current method to do that is classifying texts into positive and negative according to the sentiment orientation in the texts, but can not further measure the degree of the opinion orientation. This paper presented a quantitative method to compute the degree of sentiment orientation in documents. The algorithms took a set of documents on a topic as input and return a real number to indicate the degree of the opinion expressed in the set of documents. Experiments indicate that the algorithms work efficiently and that the ratings they produced match human ratings closely.