Financial sentiment analysis has emerged as a crucial tool for deciphering the complex narratives within financial markets. However, the accuracy of sentiment attribution to specific financial entities remains a significant challenge. This study introduces a novel framework that integrates state-of-the-art coreference resolution (CR) with a robust causal inference methodology—propensity score matching (PSM) and difference-in-differences (DID)—to precisely measure the impact of entity-specific sentiment. Coreference resolution is employed to accurately attribute sentiments to financial entities, significantly enhancing the precision of the sentiment signal. By leveraging a financial domain-adapted language model (FinBERT) and a comprehensive analytical framework, we demonstrate the causal impact of CR-enhanced sentiment on key financial metrics. To address the "black-box" nature of predictive models, we further employ SHapley Additive exPlanations (SHAP) to interpret our Random Forest sentiment predictions, identifying the primary drivers of market sentiment. The findings contribute to a deeper understanding of sentiment-market dynamics, revealing the nuanced, causal impact of sentiment on market behavior. The implications of this study extend beyond academic contributions, offering valuable, interpretable insights for investment strategies, risk management, and regulatory oversight. This research represents a significant step forward in applying advanced, data-driven, and interpretable techniques to financial decision-making.
Citation: Ruicheng Liu, Erik Cambria. Integrating financial sentiment analysis with coreference resolution: A comprehensive empirical framework[J]. Data Science in Finance and Economics, 2025, 5(4): 577-600. doi: 10.3934/DSFE.2025023
Financial sentiment analysis has emerged as a crucial tool for deciphering the complex narratives within financial markets. However, the accuracy of sentiment attribution to specific financial entities remains a significant challenge. This study introduces a novel framework that integrates state-of-the-art coreference resolution (CR) with a robust causal inference methodology—propensity score matching (PSM) and difference-in-differences (DID)—to precisely measure the impact of entity-specific sentiment. Coreference resolution is employed to accurately attribute sentiments to financial entities, significantly enhancing the precision of the sentiment signal. By leveraging a financial domain-adapted language model (FinBERT) and a comprehensive analytical framework, we demonstrate the causal impact of CR-enhanced sentiment on key financial metrics. To address the "black-box" nature of predictive models, we further employ SHapley Additive exPlanations (SHAP) to interpret our Random Forest sentiment predictions, identifying the primary drivers of market sentiment. The findings contribute to a deeper understanding of sentiment-market dynamics, revealing the nuanced, causal impact of sentiment on market behavior. The implications of this study extend beyond academic contributions, offering valuable, interpretable insights for investment strategies, risk management, and regulatory oversight. This research represents a significant step forward in applying advanced, data-driven, and interpretable techniques to financial decision-making.
| [1] |
Bohnet B, Alberti C, Collins M (2023) Coreference resolution through a seq2seq transition-based system. T Assoc Comput Ling 11: 212–226. https://doi.org/10.1162/tacl_a_00543 doi: 10.1162/tacl_a_00543
|
| [2] | Card D, Krueger AB (1994) Minimum wages and employment: A case study of the fast-food industry in new jersey and pennsylvania. Am Econ Rev 84: 772–793. |
| [3] | Clark K, Manning CD (2016) Deep reinforcement learning for mention-ranking coreference models. In: Jian Su, Kevin Duh, and Xavier Carreras, editors, Proceedings of the 2016 Conference on Empirical Methods in Natural Language Processing, 2256–2262, Austin, Texas. Association for Computational Linguistics. https://doi.org/10.18653/v1/D16-1245 |
| [4] | Devlin J, Chang MW, Lee K, et al. (2019) BERT: Pre-training of deep bidirectional transformers for language understanding. In: Jill Burstein, Christy Doran, and Thamar Solorio, editors, Proceedings of the 2019 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, Volume 1, 4171–4186, Minneapolis, Minnesota. Association for Computational Linguistics. https://doi.org/10.18653/v1/N19-1423 |
| [5] | Dobrovolskii V (2021) Word-level coreference resolution. In: Proceedings of the 2021 Conference on Empirical Methods in Natural Language Processing, 7670–7675, Punta Cana, Dominican Republic. Association for Computational Linguistics. https://doi.org/10.18653/v1/2021.emnlp-main.605 |
| [6] |
Du K, Zhao Y, Mao R, et al. (2025) Natural language processing in finance: A survey. Inform Fusion 115: 102755. https://doi.org/10.1016/j.inffus.2024.102755 doi: 10.1016/j.inffus.2024.102755
|
| [7] |
Fama EF, French KR (1993) Common risk factors in the returns on stocks and bonds. J Financ Econ 33: 3–56. https://doi.org/10.1016/0304-405X(93)90023-5 doi: 10.1016/0304-405X(93)90023-5
|
| [8] |
Gupta A, Dengre V, Kheruwala HA, et al. (2020) Comprehensive review of text-mining applications in finance. Financ Innov 6: 1–25. https://doi.org/10.1186/s40854-020-00205-1 doi: 10.1186/s40854-020-00205-1
|
| [9] | Hovy E, Marcus M, Palmer M, et al. (2006) OntoNotes: The 90% solution. In: Robert C. Moore, Jeff Bilmes, Jennifer Chu-Carroll, and Mark Sanderson, editors, Proceedings of the Human Language Technology Conference of the NAACL, Companion Volume: Short Papers, 57–60, New York, USA. Association for Computational Linguistics. Available from: https://aclanthology.org/N06-2015. |
| [10] | Jiang J, Chen J, Gu T, et al. Anomaly detection with graph convolutional networks for insider threat and fraud detection. In: MILCOM 2019-2019 IEEE Military Communications Conference (MILCOM), 109–114. IEEE. https://doi.org/10.1109/MILCOM47813.2019.9020760 |
| [11] |
Joshi M, Chen D, Liu Y, et al. (2020) SpanBERT: Improving pre-training by representing and predicting spans. T Assoc Comput Ling 8: 64–77. https://doi.org/10.1162/tacl_a_00300. doi: 10.1162/tacl_a_00300
|
| [12] | Liu Z, Huang D, Huang K, et al. (2020) Finbert: A pre-trained financial language representation model for financial text mining. In: Christian Bessiere, editor, Proceedings of the Twenty-Ninth International Joint Conference on Artificial Intelligence, IJCAI-20, 4513–4519. https://doi.org/10.24963/ijcai.2020/622 |
| [13] |
Loughran T, McDonald B (2011) When is a liability not a liability? textual analysis, dictionaries, and 10-ks. J Financ 66: 35–65. https://doi.org/10.1111/j.1540-6261.2010.01625.x doi: 10.1111/j.1540-6261.2010.01625.x
|
| [14] |
Loughran T, McDonald B (2016) Textual analysis in accounting and finance: A survey. J Account Res 54: 1187–1230. https://doi.org/10.1111/1475-679X.12123 doi: 10.1111/1475-679X.12123
|
| [15] | Peters ME, Neumann M, Iyyer M, et al. (2018) Deep contextualized word representations. In: Marilyn Walker, Heng Ji, and Amanda Stent, editors, Proceedings of the 2018 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, 1: 2227–2237, New Orleans, Louisiana. Association for Computational Linguistics. https://doi.org/10.18653/v1/N18-1202 |
| [16] | Press O, Wolf L (2017) Using the output embedding to improve language models. In: Proceedings of the 15th Conference of the European Chapter of the Association for Computational Linguistics: Volume 2, Short Papers, 157–163, Valencia, Spain. Association for Computational Linguistics. Available from: https://aclanthology.org/E17-2025. |
| [17] |
Rosenbaum PR, Rubin DB (1983) The central role of the propensity score in observational studies for causal effects. Biometrika 70: 41–55. https://doi.org/10.1093/biomet/70.1.41 doi: 10.1093/biomet/70.1.41
|
| [18] |
Tetlock PC, Saar‐Tsechansky M, Macskassy S (2008) More than words: Quantifying language to measure firms' fundamentals. J Financ 63: 1437–1467. https://doi.org/10.1111/j.1540-6261.2008.01362.x doi: 10.1111/j.1540-6261.2008.01362.x
|
| [19] | Wu W, Wang F, Yuan A, et al. (2020) CorefQA: Coreference resolution as query-based span prediction. In: Dan Jurafsky, Joyce Chai, Natalie Schluter, and Joel Tetreault, editors, Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics, 6953–6963. Association for Computational Linguistics. https://doi.org/10.18653/v1/2020.acl-main.622 |
| [20] |
Xing FZ, Cambria E, Welsch RE (2018) Natural language based financial forecasting: a survey. Artif Intell Rev 50: 49–73. https://doi.org/10.1007/s10462-017-9588-9 doi: 10.1007/s10462-017-9588-9
|
| [21] | Zerveas G, Jayaraman S, Patel D, et al. (2021) A transformer-based framework for multivariate time series representation learning. In: Proceedings of the 27th ACM SIGKDD conference on knowledge discovery & data mining, 2114–2124. https://doi.org/10.1145/3447548.3467401 |
Figures(1) / Tables(14)
Ruicheng Liu, Erik Cambria. Integrating financial sentiment analysis with coreference resolution: A comprehensive empirical framework[J]. Data Science in Finance and Economics, 2025, 5(4): 577-600. doi: 10.3934/DSFE.2025023
DownLoad: