Monetizing Currency Pair Sentiments through LLM Explainability

TL;DR

The paper addresses predicting currency-pair price movements from news-derived sentiment while maintaining interpretability. It introduces a post-hoc, LLM-based SA explainability method that identifies a $k$-sufficient set of terms $s_k$ satisfying $M(s_k)=M(T)$, and uses these explanations to enrich time-series inputs for forecasting. Empirical results on five currency pairs show that sentiment signals alone often do not improve trend accuracy, but explanations and $k$-sufficient keyword enrichment—especially with GPT-4—can boost predictive performance. The work demonstrates a practical monetary value for explainability-driven input enrichment and suggests broad generalizability to other domains and time-series forecasting tasks. This contributes a formal mechanism for extracting concise explanatory keywords from narratives and integrating them into predictive pipelines.

Abstract

Large language models (LLMs) play a vital role in almost every domain in today's organizations. In the context of this work, we highlight the use of LLMs for sentiment analysis (SA) and explainability. Specifically, we contribute a novel technique to leverage LLMs as a post-hoc model-independent tool for the explainability of SA. We applied our technique in the financial domain for currency-pair price predictions using open news feed data merged with market prices. Our application shows that the developed technique is not only a viable alternative to using conventional eXplainable AI but can also be fed back to enrich the input to the machine learning (ML) model to better predict future currency-pair values. We envision our results could be generalized to employing explainability as a conventional enrichment for ML input for better ML predictions in general.

Figures (3)

• Figure 1: Overall approach
• Figure 2: Explanation of SA with k-sufficient terms
• Figure 3: A test set example: 14 consecutive days for testing trend predictions. Red dots denote a drop and green dots denote a rise in currency-pair price with respect to the previous day.)