Systematic Literature Review of Machine Learning Models and Applications for Text Recognition

Nuzhat Khan; Ab Al-Hadi Ab Rahman; Shahriyar Masud Rizvi; Ibrahim Yousef Alshareef; Muhammad Nadzir Marsono; Muhammad Paend Bakht; Mohd Shahrizal Rusli; Shahidatul Sadiah

cs.CVAug 27, 2026

Systematic Literature Review of Machine Learning Models and Applications for Text Recognition

Authors: Nuzhat Khan, Ab Al-Hadi Ab Rahman, Shahriyar Masud Rizvi, Ibrahim Yousef Alshareef, Muhammad Nadzir Marsono, Muhammad Paend Bakht, Mohd Shahrizal Rusli, Shahidatul Sadiah

Organizations: Faculty of Electrical Engineering, Universiti Teknologi Malaysia, Johor Bahru 81310, Malaysia · Department of Electrical and Electronic Engineering, American International University-Bangladesh, 408/1, Kuratoli 1229, Bangladesh · Department of Electrical Engineering, Balochistan University of Information Technology, Engineering and Management Sciences, Quetta 87300, Pakistan · Faculty of Artificial Intelligence, Universiti Teknologi Malaysia, Kuala Lumpur 54100, Malaysia

Abstract

Optical Character Recognition (OCR) for text recognition using machine vision has significantly improved, particularly when handling heterogeneous textual data. Traditional OCR models struggle with script variations, writing styles, and degraded documents. Advancements in technology are leading to new AI models with improved architecture for handling multiple languages and complex data formats. Despite this progress, a comprehensive evaluation of OCR advancements remains limited. Based on the established preferred reporting items for systematic reviews and meta-analysis (PRISMA) guidelines, this literature review presents an extensive assessment of OCR research to trace the evolution of AI models over the past decade. It explores the transition in AI models, application domains, data types, linguistic coverage, and challenges. Through a detailed analysis of 97 selected studies published during January 2015 - January 2025, key OCR models are identified, and their performance, strengths, and limitations are analyzed. The findings highlight how OCR technologies have evolved to address structured and unstructured text, scene text recognition, and multilingual processing. Unresolved challenges include limited resources for underrepresented languages, high variability in handwritten text, visual similarity among characters, and constraints in real-time OCR applications. To address these issues, several promising approaches are proposed. Key suggestions include self-supervised learning, multimodal AI, automated machine learning (AutoML), AI-assisted postprocessing, tiny machine learning (TinyML), and the creation of joint corpora for script matching. The future recommendations aim to enhance OCR accuracy and tackle the challenges identified for real-time industrial applications. This study will guide future research and establish a foundation for OCR field.

Figures & tables

Fig. 1: Example of end-to-end car plate OCR.

Fig. 2: Review protocol.

Criteria Type	Description
Period	Focus on recent advancements in OCR. Inclusion: Publications from January 2015 to January 2025. Exclusion: Articles published before 2015.
Language	Only English language articles are included. Exclusion: The articles published in languages other than English.
Type of literature	Only high-quality, peer-reviewed sources are considered. Exclusion: Grey literature such as reports, government documents, working papers, and non-peer-reviewed sources.
Type of source	Only peer-reviewed journal and conference papers are included. Exclusion: Books, preprints, and non-peer-reviewed articles.
Accessibility	Studies must be accessible from major academic databases (IEEE Xplore, Scopus, Web of Science, SpringerLink, ScienceDirect, ACM Digital Library). Exclusion: Articles behind paywalls without institutional access.
Relevance to RQs	Articles must contribute to at least two predefined research questions. Exclusion: Articles that do not address at least two research questions.

TABLE I: Inclusion and exclusion criteria

Source	Initial Results	Retrieved Articles
Web of Science	1923	870
Scopus	9039	685
Google Scholar	21100	40
IEEE Xplore	10149	689
SpringerLink	2480	509
ScienceDirect	3121	88

TABLE II: Studies retrieved from each database in automated search

Fig. 3: PRISMA flow diagram of reviewed literature.

Fig. 4: Database distribution of the final 97 retrieved articles after systematic screening.

Fig. 5: Classification of text types in OCR.

Fig. 6: Prominent AI models applied to Optical Character Recognition (OCR) over the past decade.

Fig. 7: Evolution of OCR models architectures in a decade.

Fig. 8: Popular OCR application domains. General domains (blue box) initially focused on document digitization (green box) and expanded to other domains (orange boxes).

Model Type	Architecture	Datasets Used	Text Type / Language(s)	Performance (Accuracy)	Strength / Limitation
SVM [ 73 ] , [ 1 ] , [ 74 ] , [ 75 ] ,	SVM-MLP-ETC, NCC, HOG-SVM	286 Font Packs, Bangla Handwritten Dataset	Printed / Handwritten (English, Bangla)	60.6% - 94.53%	Works well for structured fonts but struggles with handwritten text and complex scripts
ANN [ 76 ] , [ 77 ] , [ 78 , 79 ]	Feedforward NN, Backpropagation NN, Hopfield-ANN	MNIST, CROHME, EMNIST, Picture Editor Dataset	Printed / Handwritten (English, Arabic)	92% - 100%	Effective in basic OCR but lacks generalization for large-scale real-world datasets
Neural Fuzzy System [ 80 , 81 ]	IT2NFS-SPSA, ANFIS	Alphabetic and Numeric Datasets, Burmese OCR	Printed (English, Burmese)	7.25% - 98.20%	Reduces computational cost compared to traditional fuzzy systems but limited in scalability
CNN [ 82 , 83 , 84 , 50 , 31 , 85 ]	VGG16, CRNN, Inception V3, ResNet50, Multi-CNN	MJSynth, Kaggle, IAM, BanglaLekha-Isolated, CMATERdb, EMNIST	Printed/Handwritten (English, Arabic, Tamil, Bangla, Gujarati, Devanagari)	57.2% - 99.76%	Strong for printed OCR, but struggles with handwritten cursive text and script complexity
LSTM/BLSTM [ 51 ] , [ 14 ] , [ 86 ] , [ 87 ] , [ 88 ] , [ 89 ] , [ 90 ] , [ 91 ] , [ 92 ] , [ 93 ]	Bi-LSTM, Gated-CNN-BLSTM, Tesseract LSTM	Medical Records, EMNIST, IAM, BCSD	Long-Sequenced Text (English, Arabic, German, Indonesian)	91% - 99.94%	Worked well for long-sequence text, but require high computational resources
Transformer [ 94 , 53 , 84 , 23 ]	Vision Transformer (ViT), TrOCR, CharFormer	Shotor, Sadri, Born-Digital, LicenseNet	Multilingual (Chinese, English, Persian)	77.25% - 99.98%	Outperforms CNNs/LSTMs in multilingual OCR, but computationally expensive

TABLE III: Reported accuracy of each OCR model on various datasets from selected studies

Fig. 9: Trends in language-specific OCR models.

Fig. 10: Sample of printed scripts: (a) Latin, (b) Devanagari, (c) Chinese, and (d) Arabic.

Fig. 11: A word cloud of keywords extracted from literature phrases describing challenges.

Fig. 12: Main sources of challenges in OCR.

Fig. 13: Key challenges and limitations in OCR.

Fig. 14: Data related challenges in OCR.

Model	Application	Challenges Reported	References
CNN	Multiple studies on various languages (e.g., Arabic, Bangla, Telugu, Tamil, Persian, Urdu, Gujarati, Devanagari, Javanese and Swedish)	Small text size, blurry images, non-horizontal text, complex character shapes, compound characters, diacritics, connected characters, font variations, low-quality images, and overlapping text.	[ 82 , 123 , 85 , 124 , 125 , 4 , 16 ]
DNN	Studies on OCR for different languages and document types	High computational cost, large feature matrices, memory inefficiency, data imbalance, and difficulty in recognizing handwritten text.	[ 126 , 17 , 127 ]
LSTM	Studies on handwritten OCR, license plate recognition, and document processing	Similar character shapes, joint characters, varying orientations, illumination conditions, font and background variations, and computational resource limitations.	[ 128 , 129 , 130 , 90 , 89 , 88 , 51 , 131 , 132 , 133 , 134 , 135 ]
SVM	Printed Arabic OCR and Bangla OCR studies	Feature extraction challenges, low contrast images, and difficulty handling multiple font types.	[ 136 , 73 , 137 ]
ANN	Language-oriented OCR implementations for Arabic, Persian, and Nepali scripts	Image quality issues, font variations, low resolution, and complex background interference.	[ 138 , 77 , 139 , 140 , 141 , 142 ]
Hybrid Models (CNN + LSTM, SVM + MLP, etc.)	Studies assessed ensemble approaches for OCR	Overfitting, increased training time, high memory requirements, and difficulty optimizing network layers.	[ 84 , 125 , 96 , 143 , 54 , 23 , 144 , 145 ]

TABLE IV: Challenges associated with OCR models

Fig. 15: Frequently reported challenges, existing solutions, their limitations, and future directions to overcome these issues.

Fig. 16: Recent advancements in integrated systems for OCR.

Enhancement Capability	Bilateral Filter	CLAHE	Otsu Thresholding	Gaussian Blur	Unsharp Mask	Morph. Ops	Wavelet Transf.	GAN Methods
Noise Reduction	✓	✗	✗	◦	✗	✗	✓	✓
Contrast Enhancement	✗	✓	✗	◦	✗	✗	✗	✓
Resolution Enhancement	✗	✗	✗	✗	✗	✗	✗	✓
Blur Removal	✗	✗	✗	◦	✓	✗	✗	✓
Artifact Removal	✗	✗	✗	✗	✗	✗	✗	✓
Illumination Correction	✗	✓	✗	✗	✗	✗	✗	✓

TABLE V: Comparison of traditional image enhancement techniques and GAN-based methods. (✓= Supported, ✗= Not supported, ◦= Limited support)

Explore similar work

Mar 25, 2026

PP-OCRv5: A Specialized 5M-Parameter Model Rivaling Billion-Parameter Vision-Language Models on OCR Tasks

The advent of "OCR 2.0" and large-scale vision-language models (VLMs) has set new benchmarks in text recognition. However, these unified architectures often come with significant computational demands, challenges in precise text localization within complex layouts, and a propensity for textual hallucinations. Revisiting the prevailing notion that model scale is the sole path to high accuracy, this paper introduces PP-OCRv5, a meticulously optimized, lightweight OCR system with merely 5 million parameters. We demonstrate that PP-OCRv5 achieves performance competitive with many billion-parameter VLMs on standard OCR benchmarks, while offering superior localization precision and reduced hallucinations. The cornerstone of our success lies not in architectural expansion but in a data-centric investigation. We systematically dissect the role of training data by quantifying three critical dimensions: data difficulty, data accuracy, and data diversity. Our extensive experiments reveal that with a sufficient volume of high-quality, accurately labeled, and diverse data, the performance ceiling for traditional, efficient two-stage OCR pipelines is far higher than commonly assumed. This work provides compelling evidence for the viability of lightweight, specialized models in the large-model era and offers practical insights into data curation for OCR. The source code and models are publicly available at https://github.com/PaddlePaddle/PaddleOCR.

Cheng Cui, Yubo Zhang, Ting Sun +11

PaddlePaddle Team, Baidu Inc.

Optical Character Recognition Training Data Curation

Ask PDF

Apr 14, 2026

GlotOCR Bench: OCR Models Still Struggle Beyond a Handful of Unicode Scripts

Optical character recognition (OCR) has advanced rapidly with the rise of vision-language models, yet evaluation has remained concentrated on a small cluster of high- and mid-resource scripts. We introduce GlotOCR Bench, a comprehensive benchmark evaluating OCR generalization across 100+ Unicode scripts. Our benchmark comprises clean and degraded image variants rendered from real multilingual texts. Images are rendered using fonts from the Google Fonts repository, shaped with HarfBuzz and rasterized with FreeType, supporting both LTR and RTL scripts. Samples of rendered images were manually reviewed to verify correct rendering across all scripts. We evaluate a broad suite of open-weight and proprietary vision-language models and find that most perform well on fewer than ten scripts, and even the strongest frontier models fail to generalize beyond thirty scripts. Performance broadly tracks script-level pretraining coverage, suggesting that current OCR systems rely on language model pretraining as much as on visual recognition. Models confronted with unfamiliar scripts either produce random noise or hallucinate characters from similar scripts they already know. We release the benchmark and pipeline for reproducibility. Pipeline Code: https://github.com/cisnlp/glotocr-bench, Benchmark: https://hf.co/datasets/cis-lmu/glotocr-bench.

Amir Hossein Kargaran, Nafiseh Nikeghbal, Jana Diesner +2

LMU Munich · MCML · TU Munich +1

Optical Character Recognition Multilingual VLM Evaluation

Ask PDF

May 5, 2026

CC-OCR V2: Benchmarking Large Multimodal Models for Literacy in Real-world Document Processing

Large Multimodal Models (LMMs) have recently shown strong performance on Optical Character Recognition (OCR) tasks, demonstrating their promising capability in document literacy. However, their effectiveness in real-world applications remains underexplored, as existing benchmarks adopt task scopes misaligned with practical applications and assume homogeneous acquisition conditions. To address this gap, we introduce CC-OCR V2, a comprehensive and challenging OCR benchmark tailored to real-world document processing. CC-OCR V2 focuses on practical enterprise document processing tasks and incorporates hard and corner cases that are critical yet underrepresented in prior benchmarks, covering 5 major OCR-centric tracks: text recognition, document parsing, document grounding, key information extraction, and document question answering, comprising 7,093 high-difficulty samples. Extensive experiments on 14 advanced LMMs reveal that current models fall short of real-world application requirements. Even state-of-the-art LMMs exhibit substantial performance degradation across diverse tasks and scenarios. These findings reveal a significant gap between performance on current benchmarks and effectiveness in real-world applications. We release the full dataset and evaluation toolkit at https://github.com/eioss/CC-OCR-V2.

Zhipeng Xu, Junhao Ji, Zulong Chen +10

Alibaba Group · Northeastern University

Document Understanding Multimodal Large Language Models Optical Character Recognition Multimodal Model Evaluation Document Image Understanding

Ask PDF