Character recognition by associative completion on words

Proposes a new PDAI&CD architecture and applies it to the problem of character recognition. Association is one of the most primitive functions of inference and learning from the dynamically changing outer world. PDAI&CD is just based on such a free-association function and on restrictions for ambiguity resolutions derived from strict knowledge stored as symbolic logic. As an example of PDAI&CD architecture, the system consists of an ART type neural network and of a logical word completion system which uses a spell checker. The experimental results obtained are shown here.