Enhancing Financial Market Sentiment Analysis Using Dataset Augmentation

We developed a series of financial social media sentiment analysis models starting with PyFin-Sentiment, a recent state-of-the-art domain-specific large language model for the financial domain. We focused on label accuracy and improvements in the quality and quantity of the dataset by collecting more data. We were able to improve model accuracy from 0.70 to 0.93, supporting a well-known maxim that models are just as good as the data they are trained on. This created a consistently accurate model that outperformed other existing financial sentiment models. We also used nuanced approaches such as separating social media data into short and long comments which drastically improved the accuracy of our model.