Apparatus and method for analyzing image identifications generated by an OCR device

A method for analyzing image identifications, said method comprising:

• receiving an optical character recognition (OCR) identification of an image, a confidence value associated with the received OCR identification, and a photograph from which the received OCR identification was derived;
• determining that the received confidence value is below a predefined threshold;
• after said determining that the received confidence value is below the predefined threshold, searching a data store to find at least one OCR identification stored in the data store that matches the received OCR identification, wherein the data store stores photographs of images, OCR identifications derived from the stored photographs, and correct identifications of the images in the stored photographs, and wherein the correct identifications were previously determined by a manual agent from said photographs of images;
• if said searching fails to find the at least one OCR identification, then sending a message to a manual agent, wherein the message comprises the received photograph and a request that the manual agent make a correct identification of the image in the received photograph;
• if said searching finds said at least one OCR identification, then:
• receiving a search score associated with each OCR identification of the least one OCR identification, wherein each search score is indicative of how close each associated OCR identification is to the received OCR identification;
• ascertaining whether one or more received search scores are above a predetermined threshold;
• if said ascertaining ascertains that one or more received search scores are not above the predetermined threshold then sending the message to the manual agent;
• if said ascertaining ascertains that one or more received search scores are above the predetermined threshold then determining the correct identification of the image associated with the received OCR identification based on the OCR identifications associated with the one or more received search scores above the predetermined threshold and transmitting the correct identification of the image associated with the received OCR identification to a billing system.