Knowledge-based character recognition

摘要: Character string recognition and identification is accomplished with a combined, multi-phase top-down bottom-up process. Characters in an applied signal are recognized process that employs knowledge source which contains information both, about the basic elements strings of signal. The source, may be derived from training corpus, includes word probabilities, di-gram statisitics relate likelihood words particular character prefixes, rewrite suggestions their costs. Higher level n-grams, such as tri-gram can also used. A mechanism provided for accepting not found base, well base.