科研之友
微信
新浪微博
Facebook
分享链接
摘要
The market opinions and expectations reflected in internet news provide timely and effective references for economic monitoring and early warning. In order to achieve a better identification and valuation of opinions and sentiments in economic news text, this paper proposed a public opinion dictionary construction method based on trend sentiment mapping. This method formed the trend seed word set by identifying the core key words reflecting economic trends. It integrated the calculation results of sentiment word correlation, and used the redesigned label propagation algorithm to obtain the mapping coefficient in order to obtained the opinion value of the sentiment word. Finally, it formed an inflation public opinion dictionary that can quantify news. We also proposed an inflation public opinion index model which considered the syntactic structure, through the process of topic matching, degree quantification and negative recognition to achieve a more accurate measurement of opinions and sentiments in economic news in a specific field. In the empirical analysis, an inflation public opinion dictionary was constructed and generated inflation public opinion indexes on eleven themes such as price, food and their subitems. Through classification test and comparative analysis with CPI, it was found that the public opinion index established based on the method in this paper was leading about 1.25 months than CPI in the long-term trend. The public opinion dictionary and index construction model proposed in this paper were scalable, and were expected to apply to other macroeconomics market. It was also the important improvement of the existing economic forecasting and early warning methods based on economic texts.
- 单位
