Data Extraction Using Documents RPA: Advancements and Challenges in Scanned Receipts OCR and Key Facts Extraction

Abstract

Scanned receipts OCR and key facts extraction (SROIE) constitute the processeses of spotting textual content from scanned receipts and extracting key texts from them and shop the extracted checks to established documents. SROIE performs important roles for lots report evaluation packages and holds terrific business potentials, however little or no studies works and advances had been posted on this area. In popularity of the technical challenges, significance and massive business potentials of SROIE, we prepared the ICDAR 2019 opposition on SROIE. In this opposition, we installation 3 tasks, namely, Scanned Receipt Text Localization (Task 1), Scanned Receipt OCR (Task 2) and Key Information Extraction from Scanned Receipts (Task 3). A new dataset with one thousand entire scanned receipt pictures and annotations is created for the opposition. The opposition opened on tenth February, 2019 and closed on fifth May, 2019. There are 29, 24 and 18 valid submissions acquired for the 3 opposition tasks, respectively. In this document we are able to provide the motivation, opposition datasets, project definition, assessment protocol, submission statistics, performance of submitted strategies and consequences evaluation. According to the extensive hobbies won thru SROIE and the wholesome number of submissions from academic, studies institutes and enterprise over one of kind countries, we agree with the opposition SROIE is successful. And it's far thrilling to examine many new thoughts and approaches are proposed for the brand new opposition project set on key facts extraction. According to the overall performance of the submissions, we agree with there may be nevertheless a huge hole at the anticipated facts extraction overall performance.