FinMarBa: A Market-Informed Dataset for Financial Sentiment Classification

2025-07-24

AI Summary

This paper addresses the suboptimal portfolio decisions arising from insufficient integration of financial news sentiment signals with traditional market indicators. Methodologically, it introduces a market-aware cross-modal reinforcement learning framework: (i) constructing the first open-source financial sentiment classification dataset; (ii) designing a lightweight LLM-driven sentiment encoder; (iii) proposing a three-tier hierarchical RL agent (base/meta/super) for multi-granularity decision aggregation; and (iv) establishing a scalable cross-modal feature fusion mechanism. The key contribution is the first deep embedding of financial text sentiment into dynamic asset allocation—enhancing strategy robustness and reproducibility. Empirical evaluation over 2018–2024 demonstrates an annualized return of 26% and a Sharpe ratio of 1.2, significantly outperforming both equal-weighted portfolios and the S&P 500 benchmark.

Technology Category

• Machine Learning: Multimodal Learning
• Natural Language Processing: Sentence-level Semantics, Textual Inference, etc.
• Multiagent Systems: Multiagent Learning

Application Category

• Economics, Online Markets and Human Computation: Uses of LLMs and GenAI for marketplace design, bidding, and strategic interactions
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

This paper presents a novel hierarchical framework for portfolio optimization, integrating lightweight Large Language Models (LLMs) with Deep Reinforcement Learning (DRL) to combine sentiment signals from financial news with traditional market indicators. Our three-tier architecture employs base RL agents to process hybrid data, meta-agents to aggregate their decisions, and a super-agent to merge decisions based on market data and sentiment analysis. Evaluated on data from 2018 to 2024, after training on 2000-2017, the framework achieves a 26% annualized return and a Sharpe ratio of 1.2, outperforming equal-weighted and S&P 500 benchmarks. Key contributions include scalable cross-modal integration, a hierarchical RL structure for enhanced stability, and open-source reproducibility.

Problem

• Integrating sentiment analysis with market data for portfolio optimization
• Developing a hierarchical RL framework for stable financial decision-making
• Achieving superior returns via cross-modal data and scalable architecture

Innovation

• Integrates LLMs with DRL for finance
• Hierarchical three-tier RL architecture
• Combines sentiment and market data