1.   Apart from the successive deletion of correct words from the input, the lattices are identical.

2.   Conclusions A probabilistic syntax processor has been developed to assist in the selection of the correct words for a text recognition system.

3.   Further Work Statistical processing provides a simple, efficient method for picking the correct words based on local neighbourhood.

4.   It is rarely the case that the correct word is always amongst the alternatives suggested by the lower recognition levels.

5.   The average ranking of the correct word is only three, and a number of words have to be hypothesized in the hope of including the correct word.

6.   This led to a large increase in the number of false positives, yet still did not ensure that the correct word was always the highest scoring.