Data Extraction For Interpretation of Purchase Orders From Wholesale Companies: An Approach Based On Optical Character Recognition

This work presents a method for extracting textual information from images and exporting it to administrative applications that interpret purchase orders. The research seeks an approach based on Optical Character Recognition (Optical Character Recognition - OCR), which is a system used to extract data from documents that formalize products, services, quantities, prices, and terms and conditions of payment. The experiments were carried out by making comparisons with different structures reaching 91% of assertiveness in the identification of the items.