A0670
Title: The power of visuals: Using social media images for financial sentiment analysis
Authors: Erik-Jan Senn - University of St. Gallen (Switzerland) [presenting]
Francesco Audrino - University of St Gallen (Switzerland)
Abstract: Financial sentiment analysis focuses mainly on text data. However, the importance of visual information from images has increased over the last decades, especially on social media. The objective is to investigate whether visual information influences the sentiment of retail investors and improves financial forecasting. The proposed sentiment model is based on visual information for stock-related posts on the social media platform StockTwits. The images are processed by a computer vision model and classified using user-labelled sentiment. The empirical analysis shows how visual sentiment impacts the classification performance of standard text-based models. In a financial forecasting application, the value of visual information is evaluated for financial variables such as realized volatility.
