Handwritten Optical Character Recognition (OCR): A Comprehensive Systematic Literature Review (SLR) (2001.00139v1)

Abstract: Given the ubiquity of handwritten documents in human transactions, Optical Character Recognition (OCR) of documents have invaluable practical worth. Optical character recognition is a science that enables to translate various types of documents or images into analyzable, editable and searchable data. During last decade, researchers have used artificial intelligence / machine learning tools to automatically analyze handwritten and printed documents in order to convert them into electronic format. The objective of this review paper is to summarize research that has been conducted on character recognition of handwritten documents and to provide research directions. In this Systematic Literature Review (SLR) we collected, synthesized and analyzed research articles on the topic of handwritten OCR (and closely related topics) which were published between year 2000 to 2018. We followed widely used electronic databases by following pre-defined review protocol. Articles were searched using keywords, forward reference searching and backward reference searching in order to search all the articles related to the topic. After carefully following study selection process 142 articles were selected for this SLR. This review article serves the purpose of presenting state of the art results and techniques on OCR and also provide research directions by highlighting research gaps.

• The paper presents a systematic review of 142 OCR research articles from 2000 to 2018, highlighting advancements in feature extraction and neural network integration.
• It emphasizes a paradigm shift from traditional methods to deep learning models, especially CNNs, in addressing the complexities of multilingual scripts.
• The review identifies key challenges including dataset gaps for underrepresented languages and the need for robust OCR systems in varied environmental conditions.

Handwritten Optical Character Recognition (OCR): A Comprehensive Systematic Literature Review

This essay explores the systematic literature review conducted on handwritten Optical Character Recognition (OCR), analyzing research efforts from 2000 to 2018. The review focuses on feature extraction, classification methodologies, datasets, and challenges in multilingual OCR systems, providing insights into possible future advancements.

Introduction to Handwritten OCR

Handwritten OCR has evolved significantly, leveraging AI and ML to translate handwritten text into digital formats. This digital transformation aids in preserving historical documents and facilitates various applications across different sectors, enhancing the accessibility and analyzability of handwritten data.

Methodology of the Systematic Literature Review

The review adhered to a structured protocol, incorporating articles from robust electronic databases. 142 research articles were meticulously selected based on inclusion criteria that emphasized the relevance and quality of publications related to OCR across different languages.

Statistical Overview

The publications analyzed were diverse in terms of language focus, methodologies, and datasets used. A notable surge in the adoption of deep learning techniques, particularly CNNs, marks a paradigm shift in OCR research.

Figure 1: Publications Over The Years. On the y-axis is the number of publications.

Classification Methods in OCR

Artificial Neural Networks (ANN)

The application of ANNs for handwritten character recognition has seen an upward trend, especially with deep learning architectures such as CNNs and RNNs. These networks have demonstrated remarkable success in recognizing complex patterns inherent in various languages.

Kernel Methods

Support Vector Machines (SVMs) and other kernel-based methods have been fundamentally significant in OCR for their ability to handle high-dimensional data efficiently, though they are gradually being overshadowed by neural network approaches in recent years.

Statistical and Structural Methods

Traditional statistical methods like $k$-NN still find utility in certain scenarios, whereas structural methods using graph-based representations offer promise in capturing spatial relationships among character components.

Dataset Analysis

Various languages have specific datasets that underpin OCR research, such as MNIST for English script, UCOM for Urdu, and IFN/ENIT for Arabic. The availability of diverse datasets enhances the robustness of OCR systems, although there is a notable gap in datasets for less commonly studied languages.

Critique and Future Directions

Challenges

While significant progress has been made, challenges persist, such as the need for robust systems capable of handling multilingual and cursive scripts in diverse environmental conditions ("text in the wild" scenarios).

Emerging Trends

The future of OCR lies in advancing neural network models to improve their generalization capabilities across different languages and writing styles. The development of comprehensive, high-quality datasets for underrepresented languages remains a critical area for future research.

Implications for AI Development

The insights from this review highlight a trajectory towards more intelligent, context-aware OCR systems that can contribute to the broader AI field by enhancing human-computer interaction through more effective text recognition technologies.

Conclusion

The systematic review encapsulates the considerable advancements in the field of handwritten OCR, underscoring the evolution from traditional statistical methods to sophisticated neural network models. While promising trends and methodologies emerge, ongoing innovation is essential to address existing challenges and explore uncharted territories in multilingual text recognition.