Interpretable Multilingual NLP Models for Financial Decision-Making in Frontier Markets: A Case Study on the Pakistan Stock Exchange

Abstract

The spread of artificial intelligence (AI) in the global financial markets at a very fast rate has changed the way investors perceive and respond to information. Although the sophisticated exchanges have adopted and continued implementing and maturing the natural language processing (NLP)-enhanced decision-support system, both the frontier and emergent territories have been greatly underserved by the local and clear analytical tools. This research paper is a development and testing of an explainable and multilingual AI system to be used specifically by the Pakistan Stock Exchange (PSX). The proposed system will process financial news of the Urdu and the English languages, and market disclosures using English and Urdu-local transformer-based sentiment analysis models. The sentiment indications are then put together with historical price and volume data to come up with short term directional confidence levels of the listed equities. Besides predictive performance, the framework has explainability mechanisms to enable users to track outputs of models to individual news events and sentiment drivers. As measured by empirical analysis based on historical PSX data, the directional forecasting accuracy is in fact measurably greater when English only baselines are applied to the data, as well as the interpretability and investor trust. The results imply that the information asymmetry in developing economies can be lowered through the locally adapted transparent AI-driven financial instruments, market participation can be democratized and decision quality can be improved among the retail investors. This study integrates explainable AI techniques in order to address transparency and trust challenges in financial decisions driven by AI within frontier markets. At the same time, it also models outputs that aid in decision making. The research generates the emerging body of literature concerning AI in finance by showing that explainable, market-specific decision-support systems have an economical and social worth in frontier markets.