Systematic Literature Review on General-Use Handwritten Document Processing Using OCR and LLM in Mobile Application
Handwritten documents remain common across industries. Despite advancements in handwriting digitization, specifically with the use of Optical Character Recognition (OCR) and Large Language Model (LLM), many stakeholders remain reluctant to adopt these technologies. Through analysis with Systematic Literature Network Analysis (SLNA) method on Scopus and Bibliometric Analysis with VOSViewer, which identifies a research gap in the intersection of key technologies for recent attempts in automating handwritten document processing; OCR and LLM, and specifically its adoption in mobile applications from 723 peer-reviewed studies (2020–2024) published in Scopus. Three persistent gaps emerge. First, fewer than 5% papers cover an end-to-end workflow that keeps OCR, information extraction, and post-processing natively on-device. Second, public cross-domain benchmarks for handwritten images with machine-readable data pair are not widely available, which hinders objective comparison across studies. Third, there is lack of handwritten document dataset with diverse scripts, pens, lighting, and capture angles, which hinders real-world reliability. The aim of this study is to spark development concerning fully native on-device processing systems for handwritten documents that can be tested accurately against large collections of documents containing diverse samples of handwriting. Thus, this study aims to answer the following research question; What approaches have been studied from 2020 to 2024 combining OCR technologies with LLM into a single, unified mobile application for versatile handwriting document scanning and processing at the device level, and what evidence is available regarding their effectiveness (accuracy & structure preservation), resource efficiency (latency, energy, memory), and real-world robustness across diverse scripts and capture conditions?
Authors:
Bernard Adhitya Kurniawan, Andrea Stevens Karnyoto, Bens Pardamean
2025 International Conference on Information Management and Technology (ICIMTech)