A Supervised Learning Framework for News-based Sentiment Scores Conditioned to Financial Returns
At Acuity Research Lab, in collaboration with Barcelona Tech (CS Dept.), we continuously seek for new ways to extract sentiment information from news and predict future behavior of assets returns. In this note we report our recent study and implementation of a novel methodology for sentiment analysis introduced by Zheng Tracy Ke, Bryan Kelly and Dacheng Xiu in their 2019 article titled “Predicting Returns with Text Data” [KKX]. In their paper, the authors present a supervised topic model, which they called "SESTM" (Sentiment Extraction via Screening and Topic Modeling), that uses correlation screening to create sentiment scores for Dow Jones Newswires financial news articles, customised to the prediction of financial returns.
Unlock your access to our academic stories
The Acuity team works together with the Polytechnic University of Catalonia (UPC). We work together to stay at the forefront of the latest technology, research and development of our environment. Complete the form below to read our latest academic story.