A Study on Information Extraction: Application to Administrative Document Images

This paper presents a study on the problem of information extraction and its application in building an information extraction system for administrative documents. The proposed end-to-end system contains three significant modules, including Text detection (TD), Optical character recognition (OCR), and Information extraction (IE). We developed the IE module by us based on two platforms, GraphSAGE and GATs. We have made many changes and improvements, such as redesigning graph modeling and node representation to match the goals and problems posed. We also elaborately studied to establish a complete information extraction system and dived into the information extraction module instead of all modules in the system. Besides that, we also built and evaluated our dataset of Vietnamese Administrative Documents Images (VADI2021).

I. Introduction

Document image understanding is essential in the data digitization and intelligent data management process. This relatively challenging research topic aims to recognize semantic information of text content automatically. The document understanding system is divided into two stages: text reading and information extraction. Specifically, text reading contains two significant modules, text detection and recognition. The information extraction stage is necessary to identify particular types of entities of text instances such as personal name and address.