Learning how to extract information from scanned documents
Name
1066344941-MIT.pdf
Description
Full printable version
Size
2 MB
Format
Adobe PDF
Checksum (MD5)
7d6f9e94982d8143a08dd19d3b7d4b80
Author(s)
Consul, Natasha
Advisor(s)
Regina Barzilay.
Date Issued
2017
Publisher
Massachusetts Institute of Technology
Abstract
In recent years, there has been a lot of interest in methodologies for extracting information from text-based documents. Specifically in the medical field, a recent challenge has been to extract information from different types of scanned medical documents, such as patient registration forms, prescription order forms, and medical history forms. The lack of structure and large variety of information across these documents makes it difficult to automate the process of retrieving data. Today, humans read the documents and manually record the key pieces of information. This thesis focuses on the process of learning how to automate information extraction from a variety of scanned medical documents from a Computer Vision standpoint. We look at two different approaches: an object-detection approach and a text-spotting approach . In each method, we attempt to extract a subset of document fields correctly. We evaluate and compare the results for solving the problem at hand.
Description
Thesis: M. Eng., Massachusetts Institute of Technology, Department of Electrical Engineering and Computer Science, 2017.
This electronic version was submitted by the student author. The certified thesis is available in the Institute Archives and Special Collections.
Cataloged from student-submitted PDF version of thesis. "September 2017."
Includes bibliographical references (pages 58-62).
Subjects
Electrical Engineering and Computer Science.
MIT Department
Massachusetts Institute of Technology. Department of Electrical Engineering and Computer Science
Terms of Use
MIT theses are protected by copyright. They may be viewed, downloaded, or printed from this source but further reproduction or distribution in any format is prohibited without written permission.
Persistent DSpace Link
