This directory contains handwriting samples for use in training and testing
symbol recognizers.  The samples have been encoded in the UNIPEN format 
and can be found in the include directory with the corresponding keys found in the
data directory. The include directory contains two training sets and one testing
set.  Note that there are no corresponding keys in the data directory for the 
testing data.  However, each file contains 12 samples for a particular symbol. 
The filename says what symbols that file contains.  For example, a.dat contains 
12 handwritten a's.  A total of 11 writers were used, each providing samples for 
48 symbols. The two training sets each contain 10 samples per symbol.

If there are any questions, you can contact

Joseph J. LaViola Jr.
Email: jjl@cs.brown.edu