A Review of Sentiment Analysis Algorithm for Financial News Using Natural Language Processing

Abstract

This paper presents an exhaustive examination of sentiment analysis methods utilized in financial news through Natural Language Processing (NLP). The work methodically analyzes lexicon-based, machine learning, and deep learning methodologies, encompassing VADER, the Loughran–McDonald dictionary, Support Vector Machines, LSTM, and transformer models like BERT. The review delineates the advantages and drawbacks of each method in assessing financial sentiment, especially with contextual comprehension, domain relevance, and computational efficacy. Research demonstrates that whereas lexicon-based approaches afford interpretability, deep learning models exhibit enhanced efficacy in managing intricate financial jargon. The research examines the incorporation of sentiment variables into stock market prediction models, highlighting their influence on enhancing predictive accuracy and directional forecasting. This review contributes to the literature by synthesizing recent advancements, identifying research gaps, and providing guidance for future studies, including multilingual sentiment analysis, real-time processing, and the incorporation of alternative data sources such as social media and ESG-related news.