In 2001, while I was an intern at the Bach-Archiv in Leipzig I developed a prototype of a Scribe Database which holds information about scribes and copyists of Bach manuscripts.

The goal of this project was to group manuscripts according to their handwriting similarity. I achieved this goal by linking the handwriting characteristics in the manuscripts to a feature tree that contained basic categories of handwriting characteristics. These categories were taken from Georg von Dadelsen’s “Beiträge zur Chronologie” a major publication in research on the handwriting of Bach and his copyists. By linking manuscripts and handwriting categories it was now possible to search for all manuscripts that contained a certain kind of handwriting characteristic.

A simple demonstration of this database is online at my harvard computer society server:

http://www.hcs.harvard.edu/~roeder/

In 2003 – 2005, a similar approach to the problem was taken by the Rostock-based group eNote History in a research project that was funded by the Deutsche Forschungsgemeinschaft (DFG).

I am currently working on a completely different approach that involves automatic recognition of handwriting characteristics with the help of neural nets. More info on this project can be found on the SCRIBE Project website at http://www.scribeproject.org.