xFiTRNN: A hybrid self attent linearized phrase structured contextualized transformer based RNN for financial sentence analysis with sentence level explainability

xFiTRNN is a hybrid model designed for financial sentence analysis, integrating the power of a self-attention mechanism, linearized phrase structures, and a contextualized transformer-based Recurrent Neural Network (RNN). The model leverages a novel architecture to enhance the understanding of financial sentences, enabling both high accuracy in sentiment classification and interpretability at the sentence level. By combining transformer-based contextualization with RNN’s sequential processing, xFiTRNN effectively captures the complex dependencies and semantic nuances in financial text. Additionally, the model provides sentence-level explainability, offering transparency in decision-making, which is crucial for applications in financial analysis and risk management.

In Scientific Nature