Explainable AI for Financial-News Sentiment Mining
Listed in
This article is not in any list yet, why not save it to one of your lists.Abstract
We introduce Explainable AI for Financial-News Sentiment Mining, a complete research workflow that anyone can reproduce at no cost. The workflow relies only on the free-tier of the Marketaux news-aggregation application programming interface (API) and widely available Python libraries. Our objective is to retrieve every United States technology-sector news article (tagged in the industries Technology and Semiconductors) published between January and June 2025, save each day’s material in a clear Excel format, and measure sentiment with methods that remain understandable to human readers. The sentiment layer combines two transparent techniques. First, we use VADER (Valence Aware Dictionary for Sentiment Reasoning), a rule-based lexicon that assigns a polarity score to each word, allowing us to highlight positive and negative tokens directly in the headline text. Second, we train a simple term-frequency–inverse-document-frequency model (TF–IDF) with logistic regression and explain its decisions through SHAP (Shapley Additive Explanations) feature-importance plots. This dual approach shows both the individual words that drive sentiment and the broader patterns that a supervised model discovers. We expect the result to be a practical, transparent toolset that helps researchers and practitioners explore news sentiment.