SBIR Phase II: Automatic Extraction of Financial Data from Text

The broader impact/commercial potential of this Small Business Innovation Research (SBIR) Phase II project will result from the availability of relevant financial information in structured computer-usable format with high accuracy in near real time. Currently, data embedded in financial text are extracted manually by hundreds of people working for data warehouses. This manual effort takes on the order of weeks making the bulk of the data unavailable in easily computer-usable form in real time. The benefits of this project will be focused in three areas: (i) algorithmic trading programs will be able to use all data published worldwide immediately after the data is published; (ii) financial data warehouses will be able to provide an order of magnitude larger set of data concepts (from<200 to>3000); (iii) there will be increased transparency in the financial markets as financial information embedded in text becomes computer-readable. In 2012 algorithmic trading was estimated to exceed $5 Trillion in value with 750 Billion shares traded, generating a profit of over $600 Million. Financial transparency is an intangible benefit that will improve financial market efficiency. This Small Business Innovative Research (SBIR) Phase II project will develop automated methods to extract and tag relevant financial concepts from the free text of financial documents such as an annual reports, press releases and analyst reports. The extracted financial concepts will be semantically mapped (tagged) to a financial taxonomy such as the US-GAAP or IFRS for standardized analysis. There is a growing need in current financial markets for accurate and timely access to relevant financial information for supporting trading and analysis decisions. At the end of this project, the company's technology is expected to have the capability to provide such information to analysts and decision makers in a timely fashion. The primary goals of this project will be to: (i) build an end-to-end prototype system for automatically extracting financial data from text; (ii) extend the scope of the technology to reach a broader range of real-world applications; (iii) increase accuracy; (iv) reduce the processing time for financial data extraction. The technology will employ machine learning and natural language processing techniques toward financial concept annotation, extraction and semantic tagging to achieve these goals.