基于趋势情感映射的物价舆情词典及舆情指数构建研究

The construction of inflation public opinion dictionary and index based on trend and sentiment mapping

互联网新闻中反映的市场观点和情感可为经济监测预警提供及时有效的参考.为更好地识别和量化文本中对于经济变动趋势的观点和情感,本文提出了一种基于趋势情感映射的舆情词典构建方法.该方法识别描述经济趋势的核心词汇形成趋势种子词集,并集成情感词相关性计算结果,利用重新设计的标签传播算法得到映射系数,获得情感词的观点值,形成能够量化新闻的物价舆情词典.本文还提出了一种考虑句法结构的物价舆情指数模型,通过主题匹配、程度量化、否定识别等过程实现对特定领域经济新闻中的观点和情绪的更为精确的度量.实证分析中构建了物价舆情词典并生成物价和食品及其分项物价等11个主题的舆情指数,通过分类检验及与CPI的对比分析,发现基于本文方法所建立的舆情指数在长期趋势上领先于CPI约1.25个月.本文所提出的舆情词典构建方法和舆情指数模型具有可扩展性,有望应用于其他宏观经济或行业市场景气分析研究,是现有基于经济文本的预测预警建模方法的重要改进.

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.