Neural Networks for Handwriting Recognition - Computational Intelligence Paradigms in Advanced Pattern Classification - page 19

Fig. 6 Complete RNN handwriting recognition system (here applied to offline Arabic data)

To make the comparisons fair, the same online and offline preprocessing was used
for both the HMM and RNN systems. In addition, the same dictionaries and lan-
guage models were used for the two systems.
For all the experiments, the task was to transcribe the text lines in the test set,
using the words in the dictionary. The basic performance measure was the word
accuracy:

 insertions + substitiut ions + delitions 

100 ⋅

 

1 −

 

number _ of _ words _ in _ transcript ion

 

where the number of word insertions, substitutions and deletions is summed over
the whole test set. For the RNN system, we also recorded the character accuracy,
defined as above except with characters instead of words.