METHOD FOR MACHINE RECOGNITION OF CHEMICAL STRUCTURES

THE LONG-TERM OBJECTIVE IS TO ENABLE CHEMISTS TO SCAN CHEMICAL STRUCTURES AND HAVE A COMPUTER PROVIDE ACCURATE AND TIMELY IDENTIFICATION OF THE STRUCTURE. THIS IS PROPOSED TO BE DONE BY SCANNING IN A CHEMICAL SYMBOL USING ANORMAL SCANNING PROCESS AND SHIPPING THE RESULTANT DIGITIZEDIMAGE TO A COMPUTER PACKAGE, THE NEOCOGNITRON. THE NEOCOGNITRON IS BASED ON A NEURAL NETWORK TECHNOLOGY THAT IS EXCELLENT AT PATTERN RECOGNITION. THE NEOCOGNITRON HAS DEMONSTRATED THE ABILITY TO RECOGNIZE HANDWRITTEN LETTERS AND NUMBERS. IT WILL BE TRAINED TO RECOGNIZE COMPLEX CHEMICAL SYMBOLS AND TO PROVIDE THE CORRECT CHEMICALREFERENCE NUMBER. THIS WILL PROVIDE ACCESS TO A CHEMICAL CATALOG THAT CONTAINS THE CHEMICAL STRUCTURE, THE NAME, AND PERTINENT INFORMATION. SCALING AND TRAINING ISSUES WILLBE ADDRESSED IN FITTING THE NEOCOGNITRON TO IDENTIFY CHEMICAL SYMBOLS.