I-Corps: Automatic Data Capture System using Machine Learning

University Of Houston, Houston TX

Investigators

Abstract

The broader impact/commercial potential of this I-Corps project is to transform existing data entry and document archiving systems by introducing an Automatic Data Capture System (ADCS) enhanced by machine learning. Today many sectors, including the mortgage industry, manufacturing, healthcare, and government agencies, are using cumbersome data entry and document processing techniques. This process not only adds substantially to cost, but also increases inaccuracies in the data pool, cutting into the competitive advantage of industries and agencies. The new system eliminates traditional methods of data extraction and document classification by using a self-learning software program that gathers information from any scanned document in any paper-intensive business processes. Implementing Robotic Process Automation in data entry business processes in a structured way can accurately extract and process data much faster at a very low cost point. In addition, the proposed project plan will enhance knowledge about and research on electronic data capture systems used in healthcare, allowing the development of more efficient data-processing techniques that can improve patient care and speed up billing cycles. This I-Corps project will implement an effective solution that eliminates time-consuming and error-prone manual data entry. The project will provide valuable insight into the process of scanning, data capture techniques, and machine learning. The development of technical skills will also be enhanced by an innovative learning framework that explores the application of Convolutional Neural Networks (CNNs) to data capture and processing. This proprietary approach to utilizing CNNs in a non-traditional manner has led to exceptional results in identifying and extracting information from scanned documents, making it possible for the first time to fully automate paper-intensive business processes. The commercial success of such a software program is validated by the efforts several companies are making to develop a similar product themselves because currently available technologies, such as Optical Character Recognition (OCR), cannot offer the level of accuracy required. This new Automated Data Capture System will also enhance research on virtual data systems for representing, querying, and automating data derivation - research that is used in applications such as virtual data language interpreters and data grids.