Initial training of receipts
Initially an OCR project is created for the different document classes and extraction fields of the customer receipts. In this OCR project all document classes and document fields are configured via phrases and/or rules according to the customer requirement. To simplify the training, business transactions are presorted into different stacks and introduced to the system. Based on the content (sentences, phrases, document structure, layout, etc.) the system is autonomously learning the respective business transactions or document classes.
Documents with a recognition rate lower than a defined threshold are sent to the training workspace. Inside the OCR Training Client, unrecognized documents are learned by simple association (clicking) of unknown values. The training is easily doable even for technically inexperienced users.
Interface for receipt training