AI-OCR technology

Digitization of handwritten documents by image analysis

Automatic Form Information Extraction free people from the tedious and error-prone task of typing

Form-like documents are essential in enterprise daily operations. Invoices, purchase orders, tax forms, insurance quotes, and enrollment forms are common in business workflows. They are often scanned into raster images for digital archiving and transmission. Retrieving structured information from these documents often demands manual data entry, which is tedious and unscalable. To improve efficiency and unlock business insights, automatic extraction from form data has a great demand. aiDocuDroid, an information extraction solution by means of advanced AI tachnology developed at KonicaMinolta can automatically extract the information, and thus ease the problem of resource shortage, relieve people from boring tasks to creative work.

Technology Overview

Form-like documents exhibit a high layout variation. The layout design can be affected by culture, customs, language, and aesthetic considerations. Conventional zonal solution is routinely used to guide data extraction from a particular form type, which can be cumbersome and time-consuming for end-users.  The fully data-driven approace needs little user effort, but can be brittle when form structural and layout deviates from their training samples, which happens frequently in practice. At KonicaMinolta, we strike a balance between robustness and convenience, by keeping using form templates from which keywords and structures are detected automatically. Our solution, named aiDocuDroid, can handle a large variety of forms with varying structures and layouts with high accuracy and minimal user effort.