What is post-processing OCR (lexicon, language model rescoring)?

Post-processing OCR is the stage of processing optical character recognition results aimed at improving the accuracy and quality of the recognized text. It includes several techniques:

• Lexicon — using a dictionary: after initial recognition, the system compares the obtained words with a predefined dictionary, correcting typos and errors that do not match acceptable words.

• Language Model and rescoring: the language model evaluates the probability of a sequence of words, helping to select the most likely option from several candidates proposed by OCR. For example, if OCR recognizes similar words, the language model can choose the one that best fits the context.

Thus, post-processing helps reduce the number of errors caused by noise, poor image quality, or complex typography, improving the overall recognition quality.