Paper: arXiv 2503.03612

Authors: Kemal Kirtac, Guido Germano

Abstract

Financial sentiment has become a crucial yet complex concept in finance, increasingly used in market forecasting and investment strategies. Despite its growing importance, there remains a need to define and understand what financial sentiment truly represents and how it can be effectively measured. We explore the nature of financial sentiment and investigate how large language models (LLMs) contribute to its estimation. We trace the evolution of sentiment measurement in finance, from market-based and lexicon-based methods to advanced natural language processing techniques. The emergence of LLMs has significantly enhanced sentiment analysis, providing deeper contextual understanding and greater accuracy in extracting sentiment from financial text. We examine how BERT-based models, such as RoBERTa and FinBERT, are optimized for structured sentiment classification, while GPT-based models, including GPT-4, OPT, and LLaMA, excel in financial text generation and real-time sentiment interpretation. A comparative analysis of bidirectional and autoregressive transformer architectures highlights their respective roles in investor sentiment analysis, algorithmic trading, and financial decision-making. By exploring what financial sentiment is and how it is estimated within LLMs, we provide insights into the growing role of AI-driven sentiment analysis in finance.

Complexity vs Empirical Score

  • Math Complexity: 3.5/10
  • Empirical Rigor: 4.0/10
  • Quadrant: Philosophers — conceptual discussion, lower math and empirical scores

Why this score: The paper reviews existing LLM architectures and sentiment constructs with minimal novel mathematical derivation, relying instead on comparative model descriptions. While it references prior empirical studies and trading strategies, the provided excerpt lacks specific code, dataset details, or implementation-focused backtesting metrics.

Research Flowchart

  flowchart TD
  A["Research Goal: Defining & Measuring<br>Financial Sentiment with LLMs"] --> B["Data/Inputs:<br>Financial Texts & Market Data"]
  
  B --> C["Evolving Methodology:<br>Market/Lexicon → NLP → LLMs"]
  
  C --> D["LLM Architectures & Analysis"]
  
  D --> E["BERT-based Models<br>RoBERTa / FinBERT<br>Structured Classification"]
  
  D --> F["GPT-based Models<br>GPT-4 / OPT / LLaMA<br>Text Generation & Real-time"]
  
  E --> G["Key Outcomes:<br>Enhanced Sentiment Accuracy<br>Contextual Understanding<br>Investor Analysis & Trading"]
  
  F --> G

Evidence Card

What the authors claim, on what data, and who has checked it. Read from the paper's own text by a language model: every quote is verbatim, nothing was re-run. How to read this card

Proposes

The paper proposes to define what financial sentiment is and to evaluate how large language models can be used to estimate it.

Source sentence
We explore the nature of financial sentiment and investigate how large language models (LLMs) contribute to its estimation. By exploring what financial sentiment is and how it is estimated within LLMs, we provide insights into the growing role of AI-driven sentiment analysis in finance.

Assumes

No explicit assumptions found in the text.

Data

No dataset named in the text.

Costs

Not applicable The paper reports no trading or execution result.

Finds 3

  • LLM-based sentiment approaches significantly outperformed traditional sentiment methods in predicting stock returns. predictive-accuracy
    Source sentence
    More recently, Kirtac and Germano (2024a) examined the role of large language models (LLMs) in financial sentiment analysis and stock return prediction, finding that LLM-based approaches significantly outperformed traditional sentiment methods.
  • LLM-driven sentiment measures significantly outperformed traditional sentiment indices in forecasting stock price movements. performance
    Source sentence
    Their study found that LLM-driven sentiment measures significantly outperformed traditional sentiment indices in forecasting stock price movements, reinforcing the growing role of deep learning in financial sentiment research (Kirtac and Germano, 2024a).
  • Deep learning models enable more accurate and dynamic sentiment scores that capture nuances of financial language and market dynamics. predictive-accuracy
    Source sentence
    These models enable more accurate and dynamic sentiment scores, capturing the nuances of financial language, economic narratives, and market dynamics.

Checked by others

Supported by

  • Deep Learning and NLP in Cryptocurrency Forecasting: Integrating Financial, Blockchain, and Social Media Data (2023)
    On the question: Do deep learning NLP models provide superior financial sentiment signals compared to dictionary‑based approaches?
    Compare the findings

    This paper: Deep learning models enable more accurate and dynamic sentiment scores that capture nuances of financial language and market dynamics.

    That paper: Deep learning NLP models outperform dictionary‑based sentiment analysis.

    Both findings state that deep learning models outperform dictionary‑based sentiment analysis, offering more accurate and dynamic sentiment scores.

Code

No repository by the authors found in the paper.

Extracted 2026-11-14 · same card as JSON · A language model chose these items: check the source sentence before relying on one.