Sentiment and Volatility in Financial Markets: A Review of BERT and GARCH Applications during Geopolitical Crises
Domenica Mino, Cillian Williamson · 2025
Read Sentiment and Volatility in Financial Markets: A Review of BERT and GARCH Applications during Geopolitical Crises by Domenica Mino, Cillian Williamson (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2510-16503.txt.
Introduction
Domenica Mino and Cillian Williamson review AI tools for market mood in their 2025 paper. They ask how war news mood links to stock market swings. Their data set holds news from big US outlets in 2024. The span runs from January 1 to July 17, on the Russia-Ukraine war.
A tuned BERT model, a text model trained on vast text, scores the mood of each piece. A GARCH model then tracks how swings bunch and fade over time. The pair find bad war news mood links to more market swings. The link held up under strict tests for chance. Blends of mood scores and swing models can aid risk work in crises.
Catalog record
| CREATOR | Domenica Mino, Cillian Williamson |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 45 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2510-16503
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Sentiment and Volatility in Financial Markets: A Review of BERT and GARCH Applications during Geopolitical Crises Domenica Mino, Cillian Williamson Abstract Artificial intelligence techniques have increasingly been applied to understand the complex relationship between public sentiment and financial market behaviour. This study explores the relationship between the sentiment of news related to the Russia-Ukraine war and the volatility of the stock market. A comprehensive dataset of news articles from major US platforms, published between January 1 and July 17 2024, was analysed using a fine-tuned Bidirectional Encoder Representations from Transformers (BERT) model adapted for financial language. We extracted sentiment scores and applied a Generalised Autoregressive Conditional Heteroscedasticity (GARCH) model, enhanced with a Student-t distribution to capture the heavy-tailed nature of financial returns data. The results reveal a statistically significant negative relationship between negative news sentiment and market stability, suggesting that pessimistic war coverage is associated with increased volatility in the S&P 500 index. This research demonstrates how artificial intelligence and natural language processing can be integrated with econometric modelling to assess real-time market dynamics, offering valuable tools for financial risk analysis during geopolitical crises. Keywords: artificial intelligence; sentiment analysis; BERT; GARCH; financial volatility; geopolitical news; NLP JEL classification: G14; G41; C58; C55 1 Introduction The influence of online news, social media, and television on public perception and investor sentiment is key to explaining the persistent fluctuations in financial markets [ 1 , 2 ] . In this context, digital media has substantial influence over societal emotions, particularly in investors’ sentiment during geopolitical crises such as the armed conflict between Russia and Ukraine, which began on February 22, 2022. Understanding the dynamics of this influence is crucial to analysing how stock markets respond during periods of heightened global tension. The literature has shown that investor sentiment significantly influences stock prices and volatility of returns [ 3 , 4 ] . However, analysing investor sentiment presents challenges due to the large volume of unstructured online data, such as news articles, social media posts and online forum discussions [ 5 , 6 ] . Traditional sentiment analysis techniques such as lexicon-based methods and basic machine learning algorithms have encountered limitations in handling complex and diverse data embedded in financial language. Recent advances in natural language processing (NLP), particularly through deep learning, have introduced powerful models capable of overcoming these limitations. An example is the Bidirectional Encoder Representations from Transformers model, which has significantly improved the capture of nuanced sentiment in text-based data [ 7 , 8 ] . These improvements have transformed the field of AI through architectures capable of extracting complex hierarchical representations from data [ 9 ] . BERT’s architecture, rooted in deep neural networks, enables it to interpret complex sentence structures and semantic relationships. This makes it particularly useful for sentiment analysis in finance, where the language can be technical, context sensitive, and subtle. This study aims to contribute to the growing knowledge of applied artificial intelligence and finance by examining how sentiment extracted from digital media influences financial markets during geopolitical instability. Specifically, it investigates the relationship between the sentiment in U.S. headlines news coverage of the Russia-Ukraine conflict and volatility in financial markets. Given the scale and complexity of unstructured textual data in digital media, this research uses the BERT model to quantify the sentiment from news articles, while a GARCH model enhanced with a Student t distribution is used to model the volatility in market returns. The structure of this paper is as follows. Section 2 reviews the literature, Section 3 details the data and methodology employed in this research, Section 4 presents the results, Section 5 discusses the findings, and Section 6 offers conclusions and recommendations for future research. 2 Literature Review and Conceptual Framework This section reviews the relevant literature on the influence of investor sentiment on financial markets, particularly during geopolitical crises. It evaluates traditional and advanced sentiment analysis techniques, focusing on their application in financial contexts, and outlines how such sentiment interacts with volatility modelling through GARCH-based approaches. The review also integrates recent studies on the Russia-Ukraine conflict to contextualise the current research. 2.1 Investor Sentiment and Financial Markets Investor sentiment has been extensively studied for its influence on stock market fluctuations. Sentiment-driven trading can lead to mispricing and temporary deviations from fundamental values, as documented by [ 1 ] . Market participants often rely on sentiment indicators to anticipate potential price movements, making sentiment analysis a critical tool in financial research. Investor emotions, such as fear and optimism, drive market trends, and understanding these psychological factors is essential for predicting asset price changes. Previous studies have highlighted the role of media coverage in shaping investor sentiment. [ 3 ] and [ 4 ] demonstrated that financial news and media play a significant role in influencing stock returns. The widespread dissemination of news through digital platforms has further amplified the impact of media-driven sentiment on market movements. However, quantifying sentiment from unstructured textual data remains a challenge, necessitating the use of sophisticated natural language processing (NLP) techniques to capture its effects accurately. [ 5 ] and [ 6 ] explored various methodologies for measuring investor sentiment using social media and financial news. Their findings suggest that leveraging computational approaches to analyse textual data can provide meaningful insights into market behaviour. Despite advancements in sentiment quantification, limitations persist due to the complexity and context-dependent nature of financial texts, necessitating further research in this domain. 2.2 Limitations of Traditional Sentiment Analysis Methods Lexicon-based sentiment analysis methods, while accessible and interpretable, are often too simplistic for financial contexts. These texts often involve domain-specific jargon, ambiguous phrasing, and implicit sentiment, making them difficult to interpret using standard techniques [ 2 , 10 ] . Hence they can misclassify sentiment in nuanced financial statements, especially when domain-specific language is involved. Machine learning models, including support vector machines (SVM) and naive Bayes classifiers, have been employed to improve sentiment classification accuracy. However, these models often require extensive labelled datasets for training, and their performance may degrade when applied to new or evolving financial texts [ 11 ] . On the other hand, Popular tools like VADER [ 12 ] have demonstrated improvements in handling informal texts, but still fall short when interpreting ambiguous financial expressions or negation-heavy news articles. These limitations reduce their predictive accuracy in high-stakes domains such as financial forecasting. Sentiment in a financial context is frequently influenced by implicit meanings, making it challenging to develop generalised models that perform well on different data sets. With the rise of social media platforms and real-time news dissemination, sentiment analysis models must adapt to the rapid influx of unstructured textual data [ 5 ] . The dynamic nature of market sentiment and the evolving language used in financial discussions underscores the need for more robust and context-aware NLP techniques. Advanced deep learning models offer promising solutions to these challenges by leveraging neural networks to better understand sentiment dynamics [ 9 ] . 2.3 Deep Learning and Transformer Models The development of deep learning and transformer-based models has revolutionised sentiment analysis. One of the most notable advances is the Bidirectional Encoder Representations from Transformers (BERT) model, which improves contextual understanding by leveraging bidirectional training [ 7 ] . Unlike traditional