Quant Report: Sentiment Analysis on Equities using Large Language Models
Jay Shah, Alexandros Taliotis
This paper explores the application of FinBERT, a financial adaptation of BERT, for sentiment analysis in financial markets. Built on the Transformer architecture, FinBERT is fine-tuned on financial texts to capture domain-specific language and sentiment nuances.
At King’s Capital, FinBERT was implemented to analyze equity news articles, providing sentiment scores to support trading strategies. Using data from CNBC and Yahoo Finance, headlines were preprocessed and analyzed, revealing predominantly positive sentiment for Apple-related news, with a significant neutral component. This study highlights the effec- tiveness of FinBERT in extracting actionable insights for financial decision-making.