Bridging the Gap Between Natural Language and Market Dynamics via High-Dimensional Representation Learning
This work aims to improve short-term stock price prediction for financial analysts by better capturing the nuances of financial news, representing an incremental improvement over existing sentiment analysis methods.
This paper addresses the limitations of scalar sentiment scores in financial forecasting by using high-dimensional FinBERT embeddings within a Transformer architecture. The integration of Siamese-optimized embeddings improved predictive accuracy for short-term stock price movements compared to scalar baselines and raw embeddings.
Traditional multi-modal financial forecasting often relies on scalar sentiment scores, which fail to capture the nuances of financial news. To address this information loss, this paper explores high-dimensional representation learning by replacing discrete polarity ratings with dense FinBERT embeddings within a Transformer-based forecasting architecture. We benchmarked various embedding strategies on the FNSPID dataset, including raw embeddings, attention-weighted aggregation, and a custom Siamese network. While the attention-based mechanism struggled with the low signal-to-noise ratio typical of financial data, the integration of Siamese-optimized embeddings outperformed both the scalar baseline and raw embedding approaches, demonstrating that preserving high-dimensional narrative context yields improved predictive accuracy for short-term stock price movements.