Building a Financial Polarity Dictionary Using Stock Price Information: Analysis and Verification in Chinese Stock Markets

Abstract

This paper proposes a method of building a polarity dictionary using news articles and stockprices in the Chinese market by textual analysis in finance. In order to measure the degree of polarity, weassociated the news articles' sparse composite document vectors to a score. The score is calculated by themethod of event study with the abnormal change rate of stock prices on the publication date. Weconducted support vector regression (SVR) and built a polarity dictionary with polarity data from learners.Furthermore, we made a comparison on accuracy to traditional ways of calculating word polarity inwhich news articles are represented by a one-hot wordlist. The comparison of the existed polarity is made.