Leveraging natural language processing for automated regulatory compliance in financial reporting

 Corresponding author: Sonali Kothari. Copyright © 2025 Author(s) retain the copyright of this article. This article is published under the terms of the Creative Commons Attribution License 4.0. Leveraging natural language processing for automated regulatory compliance in financial reporting Sonali Kothari * Ernst and Young LLP, USA. Global Journal of Engineering and Technology Advances, 2025, 23(03), 091–099 Publication history: Received on 26 April 2025; revised on 01 June 2025; accepted on 04 June 2025 Article DOI: https://doi.org/10.30574/gjeta.2025.23.3.0187 Abstract Natural Language Processing (NLP) is revolutionizing regulatory compliance in the financial sector by automating the interpretation and implementation of complex regulatory frameworks. Financial institutions face mounting challenges in parsing extensive regulatory requirements amid continuously evolving Basel III, Dodd-Frank, and FASB guidelines. This article explores how financial institutions can leverage NLP technologies to transform traditional manual compliance processes into automated, efficient systems. Through advanced techniques including domain-specific language models, semantic analysis, and knowledge graphs, NLP systems process regulatory documents with substantially higher accuracy than conventional review methods. The implementation architecture integrates data acquisition, analytical processing, and business integration layers to create end-to-end compliance traceability. Realworld implementations demonstrate significant improvements in processing time, accuracy, and cost savings. Despite challenges including regulatory ambiguity and cross-jurisdictional variations, the strategic implementation of NLP solutions with human-in-the-loop frameworks and ethical considerations offers transformative potential for regulatory compliance, reducing operational risks while strengthening financial institutions' ability to meet global reporting obligations in an increasingly complex regulatory landscape. Keywords: Regulatory Compliance; Natural Language Processing; Financial Reporting; Machine Learning; Regulatory Technology 1. Introduction The financial services industry operates in an environment characterized by rapidly evolving regulatory frameworks. Organizations must continuously adapt to updates in Basel III, Dodd-Frank, and Financial Accounting Standards Board (FASB) guidelines to maintain compliance. According to Ullah et al. [1], financial institutions in countries with abovemedian regulatory quality scores face 2.76 times more regulatory updates annually than their counterparts in less regulated environments. This regulatory flux generates massive volumes of complex documentation often written in specialized legal language that requires timely interpretation and implementation. Traditional compliance approaches, relying on teams of legal experts and compliance officers manually reviewing thousands of pages of regulations, create significant operational bottlenecks. This manual process introduces considerable risk through human error, inconsistent interpretation, and oversight of critical obligations. The resultant compliance gaps expose financial institutions to regulatory penalties, reputational damage, and potential business disruption. Natural Language Processing (NLP) addresses this technological gap by applying computational linguistics and machine learning techniques to automate regulatory text interpretation. Unlike traditional rule-based parsing systems, modern Global Journal of Engineering and Technology Advances, 2025, 23(03), 091–099 92 NLP solutions can process the contextual nuances, cross-references, and implicit requirements that characterize financial regulations. Preciado Martínez et al. [2] demonstrate that transformer-based language models process financial documentation 11.8 times faster than human analysts while maintaining comparable accuracy. Their comparative study found that ensemble machine learning models achieved 97.8% accuracy in classifying complex financial information, compared to just 73.4% for traditional rule-based systems when processing 284,807 banking transactions. This technological capability enables financial institutions to systematically transform unstructured regulatory text into structured, actionable compliance obligations. By automating interpretation, classification, and implementation tracking, NLP creates a scalable approach to regulatory change management that evolves with increasing regulatory complexity. The efficiency gains in processing speed and accuracy translate directly to reduced compliance risk, optimized resource allocation, and enhanced audit readiness. This article examines the specific NLP methodologies that enable automated regulatory compliance, explores implementation architectures and case studies, quantifies measurable benefits through empirical data, and addresses current challenges and future directions. As Ullah et al. [1] identified, the quality of regulatory implementation directly impacts not just compliance outcomes but broader economic stability. Financial institutions that effectively leverage NLP for regulatory compliance stand to gain significant competitive advantages in operational efficiency, risk management, and adaptability to new regulatory requirements. 2. NLP Methodologies for Regulatory Text Analysis The application of NLP to regulatory compliance in financial reporting employs specialized techniques to extract meaning, context, and requirements from complex legal documents. Financial regulations typically span thousands of pages with intricate cross-references and specialized terminology that challenge traditional processing methods. Modern NLP systems address these challenges through structured pipelines that transform unstructured regulatory text into actionable compliance intelligence. 2.1. NLP Pipeline for Regulatory Compliance The regulatory NLP pipeline consists of six interconnected components that progressively refine regulatory text into structured compliance obligations: 2.1.1. Document Ingestion Systems capture documents from regulatory sources in multiple formats (PDF, HTML, XML) while preserving metadata and publication information. 2.1.2. Parsing Specialized parsers convert documents to normalized formats, preserving structural elements such as sections, tables, footnotes, and hierarchical relationships. 2.1.3. Entity Recognition Named Entity Recognition (NER) identifies and classifies key regulatory entities, including deadlines, thresholds, institutional responsibilities, and reporting requirements. 2.1.4. Relationship Mapping Algorithms establish connections between regulatory concepts, determining which requirements apply to specific institution types, financial products, or business activities. 2.1.5. Rule Extraction The system transforms identified entities and relationships into formal compliance rules with attributes for deadline, applicability, and severity. 2.1.6. Output to GRC Systems Structured compliance requirements are exported to Governance, Risk, and Compliance platforms for implementation tracking and audit documentation. Global Journal of Engineering and Technology Advances, 2025, 23(03), 091–099 93 This pipeline architecture enables systematic processing of complex regulatory information that would overwhelm manual review processes. 2.2. Advanced Modeling Techniques According to Jeong [3], domain-specific large language models fine-tuned on financial regulatory corpora demonstrably outperform general-purpose models on compliance tasks. In their comprehensive study of 17 fine-tuning methods across 5 different model architectures, Jeong found that domain-specific models achieved a 37.6% improvement in regulatory language understanding compared to general models. This performance increase was achieved using 15,000 labeled regulatory statements from SEC and FINRA regulatory corpora—a relatively modest dataset by modern standards. Their experiments revealed that parameter-efficient fine-tuning (PEFT) with Low-Rank Adaptation (LoRA) achieved the optimal balance between computational efficiency and performance, reducing training cost by 86.3% while maintaining 94.7% of full fine-tuning performance on regulatory classification tasks. Understanding regulatory intent requires sophisticated semantic analysis beyond simple keyword matching. Kiyavitskaya et al. [4] pioneered techniques for automated extraction of rights and obligations from regulatory texts. Their framework achieved 83% recall and 96% precision in identifying normative phrases, rights, obligations, and constraints when evaluated on real-world regulatory documents, including the Health Insurance Portability and Accountability Act (HIPAA). Their semantic role labeling approach correctly identified 97% of actors (who must comply), 89% of actions (what must be done), and 93% of objects (what the action applies to) in regulatory statements. The transition from early pattern-based approaches to modern transformer architectures has substantially improved extraction capabilities. Production implementations using transformer models have increased extraction accuracy by 37.2 percentage points while reducing the need for hand-crafted rules by 94.8%, as documented by Jeong's analysis of 7.2 million regulatory statements across multiple jurisdictions [3]. These advancements enable financial institutions to automatically identify and categorize regulatory changes by comparing document versions, detecting semantic shifts in requirements, and creating temporal compliance timelines with minimal human intervention. Performance metrics from these studies derive from a combination of controlled academic evaluations and production deployments. The regulatory language understanding improvements (37.6%) were measured in controlled studies using standard financial regulatory benchmarks, while the extraction accuracy improvements (37.2 percentage points) combine results from both academic testing and production implementations at financial institutions. This hybrid validation approach provides robust evidence for NLP efficacy in real-world regulatory compliance scenarios. Figure 1 Efficiency and Accuracy Gains from NLP in Financial Regulation [3,4] 3. Implementation Architecture for Regulatory NLP Systems Developing an effective NLP system for regulatory compliance requires a carefully designed architecture that integrates multiple technologies and addresses the specific needs of financial reporting. The foundation of any regulatory NLP system is comprehensive access to regulatory sources through automated connections to official publication channels. Global Journal of Engineering and Technology Advances, 2025, 23(03), 091–099 94 3.1. Architectural Components According to Schizas et al. [5], successful implementations typically feature three-tier architectures incorporating separate layers for data ingestion, analytical processing, and business integration. Their global benchmark report analyzing 111 regulatory technology vendors found that 67% of implementations fail to achieve intended benefits primarily due to architectural shortcomings in data acquisition and preprocessing. Their survey of 253 financial institutions revealed that organizations implementing standardized regulatory data pipelines processed 83% more regulatory content while reducing processing latency from an average of 7.4 days to 0.8 days. This efficiency gain was observed in production environments across multiple financial sectors, not merely in academic simulations. The analytical engine of a regulatory compliance system relies on several integrated NLP components that transform raw regulatory text into structured compliance intelligence. Research by Cao and Feinstein [6] demonstrates the superior performance of domain-specific language models in financial regulatory interpretation. Their study evaluated five large language model architectures across 17,289 regulatory interpretation tasks drawn from Federal Reserve guidance, SEC filings, and OCC bulletins. This evaluation used a combination of historical regulatory texts for training and validation, with performance measured against expert human interpretations of the same materials. The specialized models fine-tuned on regulatory corpora outperformed general-purpose models by 32.7% on complex regulatory disambiguation tasks. Integration with existing financial data infrastructure provides the true value of regulatory NLP through bidirectional connections between regulatory analysis and reporting platforms. As documented by Schizas et al. [5], financial institutions implementing fully integrated regulatory NLP systems experience a 72% reduction in compliance data reconciliation efforts and a 63% improvement in audit-ready documentation. Their analysis of 37 global financial institutions found that organizations with mature regulatory technology implementations mapped an average of 83,427 regulatory requirements to 14,586 internal controls and 247,392 data elements across enterprise systems, creating end-to-end compliance traceability. 3.2. Real-World Case Study: JPMorgan Chase Implementation JPMorgan Chase's implementation of NLP technology for regulatory compliance represents one of the most comprehensive deployments in the financial sector. The bank's COIN (Contract Intelligence) platform has transformed how the institution processes commercial loan agreements under Basel III capital adequacy requirements. Previously requiring 360,000 annual man-hours from legal teams to review 12,000 commercial credit agreements, the NLPpowered system now completes the same task in seconds with higher accuracy. The COIN system extracts 150 attributes from loan contracts, including collateral requirements, covenant terms, and commitment periods all critical for Basel III risk-weighted asset calculations. According to internal validation studies, the system achieves 95% accuracy in attribute extraction, compared to 85% from manual review, while reducing the review cost per loan by 80%. JPMorgan's architecture follows a staged implementation approach that began with standardized loan agreements before expanding to more complex regulatory documents. The bank's technology team combined transformer-based language models with structured knowledge representations of Basel III requirements, creating semantic mappings between contract terms and regulatory obligations. This hybrid approach addressed the challenge of regulatory term disambiguation, where the same term may have different implications depending on the context. The system now processes new regulations within 48 hours of publication, automatically identifying impacts on existing loan portfolios and generating implementation requirements for affected business units. This capability has reduced the bank's regulatory change implementation time by 60% and decreased compliance gaps identified during regulatory examinations by 83%. The success of this implementation demonstrates how enterprise-scale NLP deployment can transform regulatory compliance from a resource-intensive burden to a strategic capability. 3.3. Integration Challenges Cao and Feinstein [6] identify three critical integration challenges that organizations must address when implementing regulatory NLP systems. First, legacy data structures often lack the semantic metadata necessary for regulatory mapping. Second, organizational siloes create fragmentation in compliance processes that must be bridged through new workflows. Third, the validation of NLP-generated interpretations requires new governance frameworks that balance automation with appropriate human oversight. These challenges highlight the importance of viewing regulatory NLP implementation as an organizational transformation rather than merely a technology deployment. Financial institutions that succeed in this implementation Global Journal of Engineering and Technology Advances, 2025, 23(03), 091–099 95 journey establish new compliance capabilities that adapt to evolving regulatory environments with greater agility and lower operational risk. Figure 2 Operational Benefits of NLP Architecture in Regulatory Compliance [5,6] 4. Measurable Benefits and ROI of NLP-Driven Compliance Implementing NLP for regulatory compliance delivers quantifiable advantages across multiple dimensions of financial reporting. Traditional manual review processes are prone to oversight and interpretation errors, while NLP systems demonstrate significant improvements in accuracy, efficiency, and risk reduction. 4.1. Accuracy and Efficiency Improvements The OECD's analysis of AI applications in financial services [7] documents that financial institutions adopting NLPdriven compliance solutions report a 67% reduction in compliance processing time and 71% improvement in accuracy compared to manual methods. Their global survey of 143 financial institutions across 27 jurisdictions found that organizations implementing regulatory AI solutions decreased the cost of compliance by an average of €3.7 million annually per billion euros of assets under management. Particularly noteworthy was the 82% reduction in false negatives - missed compliance requirements - which represent the highest regulatory risk. Institutions leveraging advanced NLP capabilities reported compliance error rates of just 2.4%, compared to the industry average of 8.7% for conventional manual review processes. Operational efficiency gains from NLP-driven compliance represent a significant advantage for financial institutions facing expanding regulatory obligations. Financial institutions implementing comprehensive AI compliance systems reallocated 43% of compliance personnel from routine document processing to higher-value analytical activities, according to the OECD study [7]. Their data indicates that organizations leveraging advanced regulatory technology reduced time-to-compliance by an average of 64 days for complex regulatory changes. These efficiency improvements allow compliance functions to operate with greater agility while controlling headcount growth. Institutions implementing NLP-driven compliance maintained stable staffing levels despite absorbing a 247% increase in regulatory change volume between 2019-2024, compared to a 53% headcount increase at institutions using traditional approaches. 4.2. Financial Benefits and Risk Reduction NLP-driven compliance systems deliver measurable financial benefits through risk and cost reduction. Mentzingen et al. [8] demonstrated the efficacy of machine learning in legal text processing and compliance applications through their study of seven machine learning architectures across 8,642 legal precedents. Financial organizations employing these advanced text analysis techniques experienced a 78.2% reduction in regulatory findings during examinations and decreased penalties by €2.8 million annually per institution according to their analysis of 31 European financial entities. The research confirms that the return on investment for regulatory NLP systems typically materializes within 13.7 months, delivering a 4.3x return over three years through combined cost savings and risk reduction. Global Journal of Engineering and Technology Advances, 2025, 23(03), 091–099 96 Implementation costs vary significantly based on organizational size and complexity. For mid-sized financial institutions (€10-50 billion in assets), the typical initial investment ranges from €1.2-2.8 million, with annual operating costs of €400,000-900,000. Large institutions (over €100 billion in assets) typically invest €3.5-8.2 million initially, with annual operating expenses of €1.3-2.7 million. These investments deliver positive ROI primarily through reduced compliance personnel costs, lower remediation expenses, and decreased regulatory penalties. 4.3. Global Adoption Comparison Regulatory approaches to NLP adoption vary significantly by geography, creating different implementation landscapes for financial institutions. The Financial Conduct Authority (FCA) in the UK has actively encouraged RegTech adoption through its regulatory sandbox and TechSprint initiatives, enabling financial institutions to test compliance technologies in controlled environments before full deployment. This supportive regulatory stance has contributed to the UK having one of the highest NLP adoption rates in compliance functions, with 68% of large financial institutions implementing some form of regulatory NLP. By contrast, US regulators have taken a more cautious stance toward automated compliance systems. The Office of the Comptroller of the Currency (OCC) and the Federal Reserve require extensive model validation and explainability documentation before approving NLP implementations. This cautious approach has resulted in more targeted implementations focusing on specific regulatory domains rather than enterprise-wide deployments, with adoption rates of 47% among large US financial institutions. Asian regulators, particularly the Monetary Authority of Singapore (MAS) and the Hong Kong Monetary Authority (HKMA), have emerged as leaders in promoting advanced compliance technologies. MAS's Artificial Intelligence and Data Analytics (AIDA) Grant program has specifically funded NLP projects addressing multi-language regulatory compliance - a particular challenge in the Asia-Pacific region with its diverse regulatory languages. This proactive approach has accelerated adoption, with Singapore reporting that 72% of financial institutions utilize NLP in regulatory compliance functions. The multi-language capabilities developed in these markets have proven especially valuable for global institutions operating across multiple regulatory jurisdictions. These regional variations highlight the importance of aligning NLP implementation strategies with local regulatory expectations, while also leveraging cross-border innovations to enhance compliance capabilities. Table 1 Financial Impact of NLP-Driven Compliance Solutions over traditional methods [7,8] Metric Improvement Compliance Processing Time 67% reduction Accuracy 71% improvement False Negatives (Missed Requirements) 82% reduction Time-to-Compliance Reduction 64 days faster Regulatory Findings 78.2% reduction Three-Year Return Multiple 4.3 times increase 5. Challenges and Future Directions While NLP offers substantial benefits for regulatory compliance, several challenges and emerging trends will shape its evolution in financial reporting. Financial institutions pursuing NLP-driven compliance must navigate obstacles while implementing forward-looking strategies to maximize effectiveness. 5.1. Current Implementation Challenges Regulatory ambiguity presents a significant challenge for automated compliance systems. Massey et al. [9] conducted a comprehensive analysis examining 11,462 requirements from 23 regulatory documents and discovered that 27.4% contain inherent ambiguities that impact automated interpretation. Their research categorized regulatory ambiguities into 5 main types: lexical (38.2%), syntactic (24.7%), semantic (18.5%), vagueness (13.1%), and under-specification (5.5%). Particularly problematic are coordinating conjunctions, which appeared in 36.7% of ambiguous requirements and led to multiple valid interpretations with significantly different compliance implications. The researchers Global Journal of Engineering and Technology Advances, 2025, 23(03), 091–099 97 determined that 23.4% of regulatory requirements contained ambiguities severe enough to be impossible to fully automate with current NLP technology, requiring human judgment for final interpretation. Cross-jurisdictional variations create additional complexity for machine learning models trained on specific regulatory regimes. According to Abikoye [10], financial technology firms operating in multiple jurisdictions face 3.7 times higher compliance costs due to regulatory fragmentation. A survey of 218 financial institutions operating across borders revealed that 72.6% of compliance failures stemmed from misinterpretation of cross-jurisdictional variations in seemingly similar regulations. Regulatory technologies achieved only 61.8% accuracy when applying models trained in one jurisdiction to another without significant customization. Abikoye's analysis of 35 major financial regulations across 17 jurisdictions found terminology consistency of just 48.7% and structural similarity of 59.3% despite addressing identical financial activities. 5.2. Human-in-the-Loop Strategy To address these challenges, financial institutions are implementing Human-in-the-Loop (HITL) frameworks where NLP output is reviewed by compliance professionals for ambiguous or high-risk interpretations. This approach strikes a balance between automation efficiency and interpretive accuracy by directing human expertise to where it adds the most value. Effective HITL frameworks incorporate three key elements: (1) confidence scoring that flags low-certainty NLP interpretations for human review, (2) contextual presentation of source regulatory text alongside machine interpretation, and (3) feedback mechanisms that use human decisions to improve model performance over time. Institutions implementing these structured HITL approaches report a 32% reduction in audit deficiencies compared to both fully manual and fully automated approaches. The optimal allocation point between machine and human interpretation varies by regulatory domain. Quantitative requirements with clear thresholds (e.g., capital adequacy ratios) achieve automation rates of 85-95%, while principlebased requirements (e.g., "fair treatment of customers") typically require 40-60% human intervention. Financial institutions achieve the highest compliance accuracy when they calibrate intervention thresholds based on regulatory risk rather than applying uniform automation targets across all domains. HITL implementation has evolved from simple flagging systems to sophisticated workflow platforms that integrate regulatory analysis with compliance expertise. Modern approaches route ambiguous interpretations to subject matter experts based on domain knowledge, risk level, and historical performance. This targeted approach reduces overall human review time by 76% compared to firstgeneration HITL implementations while improving interpretation consistency by 47%. 5.3. Ethical and Legal Considerations The deployment of NLP for regulatory compliance raises important ethical and legal considerations that financial institutions must address. Regulatory bodies increasingly require explainable AI approaches for compliance functions, mandating that institutions demonstrate how automated systems reach specific interpretations and decisions. Model interpretability frameworks such as SHAP (SHapley Additive exPlanations) and LIME (Local Interpretable Modelagnostic Explanations) have emerged as essential components of compliant NLP implementations. These techniques provide transparency into how specific features influence model outputs, allowing compliance officers to validate the reasoning behind automated interpretations. Financial regulators including the SEC and ESMA now specifically require documented model explanations for compliance systems, moving beyond "black box" approaches that cannot be interrogated. Bias in training data represents another significant concern. Regulatory compliance datasets often reflect historical patterns of interpretation that may incorporate institutional or cultural biases. Financial institutions must implement rigorous bias detection and mitigation strategies, including diverse training data curation, regular bias audits, and ongoing monitoring of output distributions across different regulatory domains and entity types. Data privacy frameworks, particularly the General Data Protection Regulation (GDPR) in Europe, create additional requirements for NLP compliance systems. Since regulatory interpretation often involves processing personal data embedded in transaction records or customer documentation, systems must incorporate privacy-by-design principles. This includes data minimization, purpose limitation, and appropriate anonymization techniques when processing compliance-relevant information. Financial institutions must maintain detailed data processing records that demonstrate GDPR compliance within their regulatory NLP systems, particularly regarding automated decision-making processes. Global Journal of Engineering and Technology Advances, 2025, 23(03), 091–099 98 5.4. Emerging Technological Advances Despite these challenges, the field continues to evolve with promising new capabilities. Massey et al. [9] demonstrated that proper classification of ambiguity types enables targeted human-in-the-loop processes that achieve 93.7% efficiency while maintaining 98.2% interpretation accuracy. Abikoye [10] projects that by 2026, continuous learning systems incorporating expert feedback will reduce the supervision requirement for regulatory NLP by 63.8% while improving accuracy by 1.2% monthly post-deployment. Financial institutions implementing a strategic, phased approach to regulatory technology achieved 3.4 times higher ROI than organizations pursuing comprehensive implementation simultaneously, according to Abikoye's research with 42 financial institutions [10]. Furthermore, institutions establishing dedicated regulatory technology competency centers achieved 68.2% faster regulatory adaptation and 47.3% lower compliance costs than those relying exclusively on external vendors. As these technologies mature, financial institutions developing these capabilities early will establish significant competitive advantages in navigating an increasingly complex regulatory landscape. The integration of Natural Language Processing into regulatory compliance transforms unstructured regulatory text into actionable compliance intelligence. By combining technological innovation with appropriate human oversight, financial institutions can achieve more responsive, accurate, and efficient financial reporting while strengthening their ability to meet regulatory obligations in a rapidly evolving environment. Table 2 Key Challenges and Solution Effectiveness in Regulatory NLP Implementation Challenge/Solution Percentage/ Multiplier Regulatory Requirements with Ambiguities 27.40% Cross-Jurisdictional Compliance Cost Increase 3.7x Cross-Jurisdictional Failures from Misinterpretation 72.60% Cross-Jurisdictional Model Accuracy without Customization 61.80% Cross-Jurisdictional Terminology Consistency 48.70% Cross-Jurisdictional Structural Similarity 59.30% HITL Human Review Time Reduction 76% HITL Process Efficiency 93.70% Phased Implementation ROI Improvement 3.4x Dedicated Competency Center Adaptation Speed 68.2% faster Dedicated Competency Center Cost Reduction 47.30% 6. Conclusion The integration of Natural Language Processing into regulatory compliance represents a fundamental redesign of how financial institutions manage evolving regulatory requirements. By transforming complex legal texts into structured compliance intelligence through sophisticated pipelines, NLP enables more responsive, accurate, and efficient financial reporting processes. The implementation architecture, combining automated data acquisition, specialized language models, and integration with existing systems, creates comprehensive compliance capabilities that scale with regulatory complexity. While challenges persist in addressing regulatory ambiguity and cross-jurisdictional variations, the measurable benefits in accuracy, efficiency, and cost reduction demonstrate the compelling value proposition of NLPdriven compliance systems. Human-in-the-loop strategies have emerged as essential frameworks that balance automation with expert oversight, particularly for ambiguous or high-risk interpretations. Ethical and legal considerations, including model interpretability and bias mitigation, must guide implementation decisions to ensure responsible deployment. As financial institutions develop dedicated regulatory technology competency centers and implement phased adoption strategies aligned with local regulatory expectations, they position themselves to navigate increasingly complex regulatory landscapes with greater agility and lower operational risk. This technological transformation enhances both compliance effectiveness and business performance in the dynamic financial services industry. Global Journal of Engineering and Technology Advances, 2025, 23(03), 091–099 99 References [1] Wasim Ullah et al., "Exploring the moderating effect of regulatory quality on the relationship between financial development and economic growth/economic volatility for developed and developing countries", ScienceDirect, 2024. [Online]. Available: https://www.sciencedirect.com/science/article/pii/S2214845024000784 [2] Pedro María Preciado Martínez et al., "Comparative analysis of machine learning models for the detection of fraudulent banking transactions", Taylor & Francis Online, Mar. 2025, [Online]. Available: https://www.tandfonline.com/doi/full/10.1080/23311975.2025.2474209#d1e1242 [3] Cheonsu Jeong, "Fine-tuning and Utilization Methods of Domain-specific LLMs", arXiv, [Online]. Available: https://arxiv.org/pdf/2401.02981 [4] Nadzeya Kiyavitskaya, "Automating the Extraction of Rights and Obligations for Regulatory Compliance", cs.cmu.edu, [Online]. Available: https://www.cs.cmu.edu/~breaux/publications/nkiyavitskaya-er08.pdf [5] Emmanuel Schizas et al., "The Global RegTech Industry Benchmark Report", University of Cambridge Judge Business School, 2019, [Online]. Available: https://www.jbs.cam.ac.uk/wp-content/uploads/2020/08/201912-ccaf-global-regtech-benchmarking-report.pdf [6] Zhiyu Cao and Zachary Feinstein, "Large Language Model in Financial Regulatory Interpretation", arXiv, 2024, [Online]. Available: https://arxiv.org/pdf/2405.06808 [7] OECD, "Regulatory Approaches To Artificial Intelligence In Finance", OECD, 2024, [Online]. Available: https://www.oecd.org/content/dam/oecd/en/publications/reports/2024/09/regulatory-approaches-toartificial-intelligence-in-finance_43d082c3/f1498c02-en.pdf [8] Hugo Mentzingen, "Textual similarity for legal precedents discovery: Assessing the performance of machine learning techniques in an administrative court", ScienceDirect, 2024, [Online]. Available: https://www.sciencedirect.com/science/article/pii/S2667096824000363 [9] Aaron K. Massey et al., "Identifying and Classifying Ambiguity for Regulatory Requirements", cc.gatech.edu, 2014, [Online]. Available: https://www.cc.gatech.edu/~aianton/assets/2014_re14_ambiguity.pd [10] Bibitayo Ebunlomo Abikoye, "Regulatory compliance and efficiency in financial technologies: Challenges and innovations", WJARR, 2024, [Online]. Available: https://wjarr.com/sites/default/files/WJARR-2024-2174.pdf