Finreg-E / Natural Language Processing for Financial Regulation

Finreg-E / 金融监管自然语言处理

• 批准号: 10004850
• 金额: $ 49.37万
• 依托单位: FINREG-E LIMITED
• 依托单位国家: 英国
• 项目类别: Feasibility Studies
• 财政年份: 2021
• 资助国家: 英国
• 起止时间: 2021 至 无数据
• 项目状态: 已结题
• 来源: https://gtr.ukri.org/projects?ref=10004850
• 关键词: Finreg; Natural; Language; Processing; Financial

FinregE is a trusted RegTech SME, with technical expertise enabling at least 80% time and cost savings through software for managing compliance with financial services laws and regulations. In collaboration with Imperial College London, FinregE are looking to deliver their next-generation platform in providing regulatory ease to global financial institutions (FIs), as well as sharing their ideas with governments/regulators to help their strategies to improve regulatory management and supervision.Financial sector regulation has grown significantly in scale and complexity since the financial crisis. Regulatory compliance is key to organisations, underpinning an organisation's duties to stakeholders/customers, helping build trust and reduce unforced errors.Non-compliance has had devastating commercial consequences; ~£243bn fines have been imposed on FIs since 2009 \[Reuters, 2017\]. In most FIs, regulatory compliance processes are too manually-demanding and time-consuming, with FIs struggling to achieve compliance with global regulations growing more complex by the day.The accepted opinion among FIs/regulators is that the application of machine learning/natural language processing (ML/NLP) technologies to Regulatory/Legal text could be straightforward and imminent. However, standard ML/NLP/Deep-Learning used to analyse text documents perform poorly in the legal/regulatory domain and are yet to be successfully achieved. NLP research-to-date has focussed on user reviews, where the language is colloquial/informal, and of shorter length than laws/regulations.With this innovation, FinregE will deliver a one-of-a-kind ML/NLP driven regulatory compliance software that quickens and improves the accuracy of regulatory compliance, making the most effective use of technology yet in this domain. FinregE's platform will provide following solutions:1. Continuous alerts on emerging regulations/laws from multiple sources;2. Automated classification/organisation of regulations/laws to identify the full network view of regulation/law in a topic classes/domains;3. Dynamic extraction of summaries of the most pertinent information requiring compliance actions;4. The ability to map extracted information against entity's internal policies/procedures/controls to automatically alert owners and identify actions required for compliance.With FinregE's intuitive and innovative compliance software FIs and regulators will be able to determine impact and plan adoption of regulation quickly. FIs will be able to effectively triage every publication/rule to extract key information requiring compliance actions, processing and implementation regulations at least 90% faster than manual processing. The project will deliver a software and framework for a universal set of regulatory topics with defined sets of rules which will keep evolving as societal change creates new areas of law across different topics and global jurisdictions. No such enterprise-wide solution currently exists.

FinregE是一家值得信赖的监管科技中小企业，拥有技术专长，通过管理金融服务法律法规合规性的软件，可节省至少80%的时间和成本。FinregE与伦敦帝国理工学院伦敦合作，希望提供下一代平台，为全球金融机构（FI）提供监管便利，并与政府/监管机构分享他们的想法，以帮助他们改善监管管理和监督的战略。自金融危机以来，金融行业监管的规模和复杂性都大幅增长。监管合规是组织的关键，是组织对利益相关者/客户的义务的基础，有助于建立信任并减少非受迫性错误。不合规已经产生了毁灭性的商业后果;自2009年以来，金融机构已被处以约2430亿英镑的罚款\[路透社，2017\]。在大多数金融机构中，监管合规流程对人工的要求太高，而且耗时太长，金融机构难以遵守日益复杂的全球法规。金融机构/监管机构普遍认为，将机器学习/自然语言处理（ML/NLP）技术应用于监管/法律的文本可能是直接且迫在眉睫的。然而，用于分析文本文档的标准ML/NLP/深度学习在法律的/监管领域表现不佳，尚未成功实现。迄今为止，NLP的研究主要集中在用户评论上，其中的语言是口语/非正式的，并且比法律/法规更短。通过这一创新，FinregE将提供一个独一无二的ML/NLP驱动的监管合规软件，加快并提高监管合规的准确性，使该领域的技术得到最有效的利用。FinregE的平台将提供以下解决方案：1.从多个来源不断提醒新的法规/法律;2.法规/法律的自动分类/组织，以识别主题类/域中的法规/法律的完整网络视图;3.动态提取需要合规行动的最相关信息的摘要;4.能够将提取的信息与实体的内部政策/程序/控制进行映射，以自动提醒所有者并确定合规所需的行动。借助FinregE直观而创新的合规软件，金融机构和监管机构将能够快速确定监管的影响并计划采用。金融机构将能够有效地对每个出版物/规则进行分类，以提取需要采取合规行动、处理和实施法规的关键信息，其速度至少比人工处理快90%。该项目将为一套通用的监管主题提供一个软件和框架，并定义了一套规则，这些规则将随着社会变革在不同主题和全球管辖区创造新的法律领域而不断发展。目前还没有这样的企业级解决方案。