SentXStock: A Multi-Tier NLP Framework for Financial Sentiment Analysis

Financial markets are strongly influenced by public sentiment expressed through news articles and social media discussions. Rapid analysis of such information is essential for investors to make informed decisions. This paper presents SentXStock, an AI-driven financial sentiment analysis framework designed to evaluate market sentiment and assist in trading decision support. The proposed system integrates a multi-tier natural language processing architecture combining machine learning models, large language models, and rule-based sentiment analysis techniques. Financial news and social media data are collected from multiple sources and processed through a layered sentiment analysis pipeline. The system uses the FinBERT model for financial sentiment classification, contextual reasoning using large language models, and a rule-based fallback mechanism using VADER sentiment scoring. SentXStock also incorporates a portfolio simulation module that dynamically adjusts investment risk based on aggregated sentiment scores derived from market data. The system generates actionable recommendations such as buy, sell, or hold decisions and visualizes market sentiment using an interactive Streamlit dashboard. The framework demonstrates how combining modern artificial intelligence techniques can provide an effective decision support system for financial sentiment analysis and automated trading insights.