A data-driven 'knowledge engine' for financial analysis

对公司的投资是经济增长和竞争力的根本驱动力，有助于促进创新和提高生产力，但研究表明，60%的并购交易实际上破坏了价值。全球并购市场价值约4万亿美元，这意味着约2.4万亿美元被浪费了。这可能是由于尽职调查数据质量差，分析不科学，分析师无法准确评估市场风险和机会，导致交易协同效应和估值被高估。财务分析由于需要从多个难以解析的异构数据源（公司网站、研究报告、财务报表、社交媒体等）中分析和提取信息而受到阻碍，这需要人工分析师筛选数千份文件并手动处理数字。缺乏复杂的工具和数据科学专业知识意味着财务分析仍然不复杂且容易出错。整个过程是时间和成本密集的，最终无法扩展到指数级增长的网络数据量。为了应对这一挑战，plurai开发了一个用于财务分析的数据科学平台的第一版（最小可行产品），使用自然语言处理、机器学习和知识图谱技术，能够生成潜在的收购目标或估计市场规模。拟议的项目将建立在这一早期成果的基础上，开发一个完全自动化和可扩展的“知识引擎”，专门用于财务分析。与传统搜索引擎（将关键字与可用的第三方结果相匹配）不同，plurai的知识引擎采用了一种计算方法，通过解析问题并生成定制答案，可以对其来源和工作原理进行审计。这允许用户回答非常小众/复杂的问题，目前还没有答案。这个项目的重点是证明完全自动化概念挖掘和知识库创建的可行性，使引擎能够在整个web规模上处理广泛的金融查询。这包括自动确定哪些来源是值得信赖的，并生成感兴趣的实体、概念和链接——从而为改进的财务分析提供完全可扩展的、破坏性的解决方案。最初的目标应用程序是公司金融行业，特别是交易发起用例，其中需求和兴趣已经建立；然而，强大的可转移性设想到任何研究/分析工作。好处包括改进、快速的投资决策，从而增加投资和优化回报；显著节省成本和时间；最终为金融市场带来稳健、科学的决策。

Investment into companies is a fundamental driver of economic growth and competitiveness, helping to foster innovation and improve productivity, however research shows that 60% of M&A deals actually destroy value. The global M&A market is valued at ~$4 trillion, meaning ~$2.4 trillion is wasted. This can be attributed to poor quality data and unscientific analysis in due diligence, with analysts unable to accurately assess market risks and opportunities, resulting in overestimated deal synergies and valuations. Financial analysis is hampered by the need to analyse and extract information from multiple, hard to parse, heterogeneous data sources (company websites, research reports, financial statements, social media, etc.), which requires human analysts to sift through thousands of documents and manually crunch numbers. A lack of sophisticated tools and data science expertise means financial analysis remains unsophisticated and error-prone. The whole process is time and cost intensive and ultimately unable to scale given exponentially increasing amounts of web data.To address this challenge, Plural AI has developed a first version (Minimal Viable Product) of a data science platform for financial analysis, using natural language processing, machine learning and knowledge graph technology, which is capable of generating potential acquisition targets or estimating the size of a market. The proposed project will build upon this early achievement to develop a fully automated and scalable 'knowledge engine' specifically for financial analysis. Unlike traditional search engines, which match keywords to available third-party results, Plural AI's knowledge engine adopts a computational approach by parsing the question and generating a bespoke answer, which can be audited for sources and workings. This allows users to answer very niche/complex questions to which answers currently do not exist.This project focusses on proving the feasibility of fully automating the concept mining and knowledge base creation, enabling the engine to handle a wide array of financial queries at full web scale. This involves automatically determining which sources are trustworthy, and generating entities, concepts, and links of interest - thus providing a fully scalable, disruptive solution for improved financial analysis.The initial target application is the corporate finance industry, in particular for deal origination use cases, where need and interest has been established; however strong transferability is envisaged to any research/analysis work. Benefits include improved, rapid investment decision making, leading to increased investments and optimised returns; significant cost and time savings; ultimately bringing robust, scientific decision-making to the finance market.