After a review of commercial sgml encoding products the University of Chicago Library decided to create an encoding program in-house. We decided that the best way to encode a finding aid was to separate it into sections. A finding aid can be conceptually split into two distinct parts, the front matter, which is information about the finding aid itself, and the container listings, which describe the actual contents of the collection.  In order to automatically markup finding aids using our program, the finding aid must first be in electronic form (Word file). This can be achieved by either scanning or re-keying the document. The front matter is marked up using an HTML form (template). The relevant information is cut-and-pasted into the template fields and then run through a cgi-script written in Python 1.5 which outputs marked-up text to a file. This part of the program takes cgi variables and marks them up, formatting them slightly and producing the first part of the finding aid. The second half of the finding aid (container listings) is marked up using a program designed especially for this project, which searches through a text file for patterns and keywords, and outputs marked-up text to a file.

When these two files are joined together, we have a completely encoded finding aid, viewable with an SGML browser. The program examines the text line by line, looking for an indicator we have inserted () to mark the beginning of relevant material. When it finds the indicator, it then starts scanning for keywords such as Folder, Box, Series, etc. Each keyword prompts an action, or subroutine. As the program can only really look for patterns, anomalies can cause problems. The program was written to do the bulk of the work, but hand editing before and after running the program may be necessary.