BondBERT: What we learn when assigning sentiment in the bond market

TL;DR

BondBERT addresses the gap in bond-specific sentiment analysis by adapting transformers to fixed-income dynamics where sentiment and prices can move inversely to macro optimism. The paper builds a 30k-article UK bond corpus, labels sentiment with GPT-4.1-nano, and fine-tunes FinBERT to create BondBERT, achieving superior alignment with bond returns and improved forecasting over FinBERT, FinGPT, and Instruct-FinGPT. Across event-based correlations, directional accuracy, and LSTM-based price forecasting on 10 UK sovereign bonds, BondBERT delivers stronger predictive power and statistically significant improvements in several instruments. This work highlights the value of domain-adaptive sentiment for financial decision-support and points to future extensions including human annotation and broader asset coverage.

Abstract

Bond markets respond differently to macroeconomic news compared to equity markets, yet most sentiment models are trained primarily on general financial or equity news data. However, bond prices often move in the opposite direction to economic optimism, making general or equity-based sentiment tools potentially misleading. We introduce BondBERT, a transformer-based language model fine-tuned on bond-specific news. BondBERT can act as the perception and reasoning component of a financial decision-support agent, providing sentiment signals that integrate with forecasting models. We propose a generalisable framework for adapting transformers to low-volatility, domain-inverse sentiment tasks by compiling and cleaning 30,000 UK bond market articles (2018-2025). BondBERT's sentiment predictions are compared against FinBERT, FinGPT, and Instruct-FinGPT using event-based correlation, up/down accuracy analyses, and LSTM forecasting across ten UK sovereign bonds. We find that BondBERT consistently produces positive correlations with bond returns, and achieves higher alignment and forecasting accuracy than the three baseline models. These results demonstrate that domain-specific sentiment adaptation better captures fixed income dynamics, bridging a gap between NLP advances and bond market analytics.

Figures (3)

• Figure 1: Distribution of price and log returns for the 10 UK instruments.
• Figure 2: Sentiment correlations with UK bond prices over a 7-day rolling window across four models.
• Figure 3: Model performance comparison, showing directional accuracy of sentiment relative to bond returns.