Construction and Application of Sentiment Lexicons in Finance

Abstract

This article proposes an approach to constructing sentiment lexicons in the financial domain. The approach takes advantages of news bulletins and a given financial variable, such as stock prices, to generate candidates of sentiment expressions by fusing the two data sources. The candidates are then filtered based on their co-occurrences with financial seed words and are subsequently expanded by analogical reasoning using distributed representation of words. Evaluative experiments on real-world news and stock price data shows that the resulting lexicons are mostly reasonable and capture the characteristics of the target financial variables. As a potential application, trading simulation is also carried out based on the resulting financial sentiment lexicons, demonstrating the utility of the lexicons.