Method of coding characters for word recognition and word recognition device using that coding

摘要

The method of coding characters for word recognition in a word recognition device comprises the steps of: reading a plurality of known words (19); detecting morphological similarities between the characters of the alphabet forming the known words (20-22); associating with each character a weight (23), by associating morphologically similar characters with similar weights; and storing the weights in a data memory of the word recognition device. To detect the morphological similarity between the characters, a confusion table is determined, by plotting data relating to the probability of confusing a character belonging to the known words with other characters; on the basis of the confusion table, a similarity graph is determined in which the characters of the alphabet which can easily be confused with one another are connected by connection lines; and on the basis of the similarity graph, an adjacency list is determined, including the characters of the alphabet arranged by degree of reciprocal similarity. The weights are then fixed in such a way that similar weights correspond to adjacent characters in the adjacency list.