Automatic Reformatting of OCR Text from Biomedical Journal Articles
Glenn Ford, Susan E. Hauser, George R. Thoma · 1999
The goal of the Medical Article Record System (MARS), being developed by the National Library of Medicine, is to reduce the manual keyboard entry of bibliographic citation fields for the MEDLINE database by automatically identifying and converting information from bitmapped images of biomedical journal article pages to ASCII data. An important element of this automatic conversion requires reformatting the title, author and affiliation fields from the output of the Optical Character Recognition (OCR) process in MARS to the formats specified by MEDLINE conventions. This paper outlines the methods developed to implement the reformatting process.