An intelligent document analysis system for evaluating corporate governance practices based on SEC required filing

Abstract

An IDA is an Intelligent Document Analysis System for Evaluating Corporate Governance Practices capable of retrieving required documentation of public companies from the Securities and Exchange Commission (SEC) and performing analysis and rating in terms of recommended corporate governance practices .

A desired IDA system must be loose coupling, cost-effective, efficient, accurate, as well as operate in real-time. Such a sophisticated system can help individual and institutional investors with evaluating individual companies’ corporate governance practices. With the techniques of analogical learning, local knowledge bases, databases, and question-dependent semantic networks, the IDA system is able to automatically evaluate the strengths, deficiencies, and risks of a company's corporate governance practices based on the documents stored in the SEC EDGAR database. (U.S. Securities and Exchange Commission 2013) A produced score reduces a complex corporate governance process and related policies into a single number which enables concerned government agencies, investors and legislators to easily review the governance characteristics of individual companies .

The present manual process not only incurs huge overhead costs, it also has the risk associated with human error and bias in their ratings. To reduce the cost, some companies outsource the job to other countries with a low labor cost. But the problems of how to timely respond to a changing world and constantly updating information remains unanswered. To address those challenges and corresponding issues, first, we developed a system with a user friendly interface and a collection of knowledge bases, databases and semantic networks, a rating engine and web portal. Then, we ran iterative tests and experiments to identify the drawbacks. Subsequently, we developed effective techniques to address the uncertainty of finding the correct answers, with the purpose of improving the system's performance and ensuring accuracy. Lastly, we investigated and implemented machine learning techniques to efficiently improve the process of the IDA System by using different methods of evaluating similarity.