This project compares the use of three language models (BERT, DistilBERT, and RoBERTa) for sentiment analysis of statements related to financial markets. Each model was fine-tuned using financial statements that were hand-annotated by financial experts. Statements were classified as either positive, negative or neutral. Following fine-tuning, each model was evaluated on a test set of data.
Financial data used for this project can be found here: https://www.kaggle.com/ankurzing/sentiment-analysis-for-financial-news