OCR and text recognition

The rich manuscript heritage of Italy, preserved in archives and libraries, is becoming increasingly accessible to a wider audience through dedicated digitization initiatives. However, the interpretation of these manuscripts often proves challenging due to several factors: the linguistic complexity of medieval Latin, the early development of vernacular languages, the continuous evolution of handwriting styles, and the extensive use of abbreviation systems devised to conserve space on costly materials such as parchment and paper. Artificial Intelligence (AI) tools can significantly boost the last step of the digitization process: transcription. In particular, the advent of Handwritten Character Recognition (HCR) technology enables recognition and processing of handwritten text. However, as with all AI tools-especially in the domain of handwritten texts and, more broadly, in the Humanities, training and fine-tuning is required. To support Digital Humanists in tailoring these powerful tools to specific needs-i.e. transcribing different handwriting styles-a Human-AI collaboration approach has been adopted to develop a collaborative web application, named HCR WORKFLOW, designed for the creation of ground-truth data for AI-based manuscript transcription. The platform is composed of a toolkit for document layout analysis based on Neural Networks for text line recognition (P2PaLA), an image Transformer encoder and an autoregressive text Transformer decoder for single-line transcription (TrOCR). This integrated system guides and assists Digital Humanists throughout the entire process-from digitization to transcription supervision. For this study, the platform was used to fine-tune TrOCR on humanistic script, and in particular to create the ground truth based on the Copialettere (Letterbooks) of Isabella d'Este and the letters addressed to her by Lucrezia Borgia. This research paper will discuss in detail the HCR WORKFLOW platform, the dataset used, the approach to create an AI-oriented transcription, and the results of the fine-tuning of the AI tool for manuscript transcription.

The shipping and logistics industry is undergoing a transformative shift driven by rapid advancements in technology. Innovations such as artificial intelligence (AI), the Internet of Things block chain, automation, and big data analytics are significantly enhancing operational efficiency and fostering innovation across the supply chain. These technologies enable realtime tracking, predictive maintenance, route optimization, and improved inventory management, thereby reducing costs, minimizing delays, and improving customer satisfaction. Additionally, digital platforms and automated systems are streamlining processes from warehousing to last-mile delivery, while also promoting sustainability through better resource utilization. This paper explores the key technological trends revolutionizing the logistics sector and evaluates their impact on productivity, agility, and long-term competitiveness. Block chain technology offers secure, tamper-proof digital records, fostering trust and reducing fraud in international trade. Automation and robotics are revolutionizing warehousing and port operations, minimizing human error, and increasing throughput. Furthermore, cloud computing and digital platforms are driving the integration of logistics networks, enhancing collaboration among stakeholders. By highlighting real-world applications and case studies, it underscores the critical role of technology in shaping the future of global shipping and logistics. I.

The rapid growth of digital data has led to the widespread creation and storage of digital images containing text. The extraction and use of textual information might be advantageous for various kinds of domains. Text detection in natural images is primarily affected by noise, blur, distortions, font variation, alignments, and orientation. Government forms, mark cards, medical records, business receipts, and other increasingly common bilingual documents have a substantial impact about precision of text detection and recognition. The paper, "Enhanced Machine Learning Methods for Robust Real-Time Classification of Bilingual Documents," emphasizes these challenges and suggests a solution that uses image enhancement techniques to improve an image's appearance and quality. Digital documents are bilingual; hence, extracting information from them is challenging since a computer can read and interpret text written in many scripts, including English and Kannada, within the same documentation. It thus extracts text from documents using optical character recognition (OCR)-based problems. Natural language processing (NLP) to classify text, super-resolution methods to synthesize a highresolution (HR) image from several lowresolution (LR) images and machine learning models EasyOcr to detect and recognize Text.

Optical character recognition (OCR) deals with the process of identification of alphabets and various scripts. Optical character recognition (OCR) is one of the trending topics of all time. Optical character recognition (OCR) is used for pattern detection and artificial intelligence. Machine learning is widely used in the field of OCR to provide good accuracy in the result. In Python, Pytesseract is an optical character recognition (OCR) tool for python. The paper starts with an introduction and brief background and history of Optical character recognition (OCR) systems. Then the various techniques of OCR systems such as optical scanning, location segmentation, pre-processing, feature extraction and post-processing. The different applications of OCR systems are highlighted next followed by the current uses of the OCR systems. The future of the Optical character recognition (OCR) systems with machine learning environment is presented.

Old Uyghur manuscripts remain largely inaccessible due to the absence of optical character recognition (OCR) systems aligned with modern scholarly practices. To address this, we present a work‑in‑progress OCR approach using fine-tuning of vision–language models (VLMs) for Old Uyghur transliteration. Specifically, LLaMA‑3.2‑11B‑Vision was fine‑tuned with 4‑bit Unsloth quantization and LoRA on 525 manually annotated pages from Radloff and Malov’s edition of Altun Yaruk Sudur, following Unicode‑based transliteration guidelines. On a held‑out test set of 16 pages from one bölük (part) of the fifth tägzinč, the approach yielded a character error rate (CER) of 5.46% and normalized edit distance (NED) of 0.286. These results demonstrate feasibility; however, limitations are evident owing to binarized scans, a single woodblock style, small training data, and lack of positional supervision. Consequently, findings provide a baseline and a reproducible path toward robust Old Uyghur OCR.

Old Uyghur manuscripts remain largely inaccessible due to the absence of optical character recognition (OCR) systems aligned with modern scholarly practices. To address this, we present a work‑in‑progress OCR approach using fine-tuning of vision–language models (VLMs) for Old Uyghur transliteration. Specifically, LLaMA‑3.2‑11B‑Vision was fine‑tuned with 4‑bit Unsloth quantization and LoRA on 525 manually annotated pages from Radloff and Malov’s edition of Altun Yaruk Sudur, following Unicode‑based transliteration guidelines. On a held‑out test set of 16 pages from one bölük (part) of the fifth tägzinč, the approach yielded a character error rate (CER) of 5.46% and normalized edit distance (NED) of 0.286. These results demonstrate feasibility; however, limitations are evident owing to binarized scans, a single woodblock style, small training data, and lack of positional supervision. Consequently, findings provide a baseline and a reproducible path toward robust Old Uyghur OCR.

Here, we introduce the Historical Kannada Handwritten Palm Leaf (HKHPL) Dataset, the pioneering collection of historical palm leaf manuscripts written in Kannada. It includes original ground-truth and binarised images. This dataset is constructed from randomly selected palm leaf manuscripts collected from various parts of Karnataka and Maharashtra, India, and is publicly available for scientific purposes. The HKHPL dataset significantly contributes to the study of ancient documents and the recognition of handwritten Kannada texts on palm leaf manuscripts, aiding important tasks such as reading characters, segmenting text lines, and analyzing layouts. This dataset addresses challenges posed by ancient manuscripts, including degradation, non-uniform writing surfaces, and handwriting variations. By providing binarised ground truth images, word-annotated images, and isolated character-annotated images, the dataset offers researchers a valuable resource for developing and testing algorithms in image preprocessing, word segmentation, and character recognition. The diverse representations of historical Kannada texts ensure robust machine learning models capable of handling various writing styles, document conditions, and content types. With the HKHPL dataset publicly available, along with its implementation code on GitHub, the creators have opened new avenues for research in digital humanities, historical linguistics, and AI applications in cultural heritage preservation. This dataset can accelerate advancements in the automated transcription and analysis of historical Kannada texts, contributing to the preservation and study of India's literary heritage.

Extracting texts from images with complex backgrounds is a major challenge today. Many existing Optical Character Recognition (OCR) systems could not handle this problem. As reported in the literature, some existing methods that can handle the problem still encounter major difficulties with extracting texts from images with sharp varying contours, touching word and skewed words from scanned documents and images with such complex backgrounds. There is, therefore, a need for new methods that could easily and efficiently extract texts from these images with complex backgrounds, which is the primary reason for this work. This study collected image data and investigated the processes involved in image processing and the techniques applied for data segmentation. It employed an adaptive thresholding algorithm to the selected images to properly segment text characters from the image's complex background. It then used Tesseract, a machine learning product, to extract the text from the image file. The images used were coloured images sourced from the internet with different formats like jpg, png, webp and different resolutions. A custom adaptive algorithm was applied to the images to unify their complex backgrounds. This algorithm leveraged on the Gaussian thresholding algorithm. The algorithm differs from the conventional Gaussian algorithm as it dynamically generated the blocksize to apply threshing to the image. This ensured that, unlike conventional image segmentation, images were processed area-wise (in pixels) as specified by the algorithm at each instance. The system was implemented using Python 3.6 programming language. Experimentation involved fifty different images with complex backgrounds. The results showed that the system was able to extract English character-based texts from images with complex backgrounds with 69.7% word-level accuracy and 81.9% character-level accuracy. The proposed method in this study proved to be more efficient as it outperformed the existing methods in terms of the character level percentage accuracy.

The use of advanced computational methods for the analysis of digitised texts is becoming increasingly popular in humanities and social science research. One such technology is Handwritten Text Recognition (HTR), which generates transcripts from digitised texts with machine learning approaches, to enable full-text search and analysis. Up to now, HTR models for Tibetan manuscripts in cursive script have not been available. This paper introduces work carried out as part of the The Dawn of Tibetan Buddhist Scholasticism (11th-13th) (TibSchol) project at the Austrian Academy of Sciences, which is utilising the Transkribus platform to explore possible solutions to automate the transcription of Tibetan cursive scripts. It presents our methodology and preliminary results along with a discussion of the limitations and potential of our current models.

Perkembangan teknologi informasi dan komunikasi telah membawa
perubahan signifikan dalam berbagai aspek kehidupan, termasuk dalam
pengelolaan administrasi di institusi pendidikan. Pengarsipan dokumen, khususnya surat menyurat, merupakan bagian krusial dari tata kelola administrasi yang baik. Di era digital, pengarsipan konvensional mulai beralih ke pengarsipan digital untuk meningkatkan efisiensi, kecepatan akses, dan keamanan data. SMK Diponegoro Karanganyar sebagai salah satu institusi pendidikan yang terus berupaya meningkatkan mutu layanannya, telah memanfaatkan aplikasi manajemen sekolah (smeduverse) untuk mendukung berbagai kegiatan operasional,
termasuk pengelolaan surat masuk. Salah satu fitur dalam aplikasi tersebut memungkinkan pengarsipan surat masuk secara digital, di mana surat fisik yang telah didisposisi difoto dan diunggah ke dalam sistem.
Namun, proses yang berjalan saat ini masih memiliki potensi untuk
dioptimalkan. Pengelolaan arsip merupakan elemen vital dalam administrasi setiap instansi atau perusahaan, termasuk institusi pendidikan seperti Sekolah Menengah Kejuruan (SMK). Arsip, baik surat masuk maupun keluar, berfungsi sebagai sumber informasi dan rekaman kegiatan yang krusial. Seiring dengan perkembangan teknologi informasi, banyak institusi mulai beralih dari sistem pengarsipan manual ke digital untuk meningkatkan efisiensi. Namun, pengelolaan arsip secara
manual, bahkan yang sudah melibatkan pemindaian atau foto, seringkali masih menimbulkan masalah seperti kesulitan dalam pencarian dokumen ketika dibutuhkan, risiko kehilangan data, dan proses yang memakan waktu. Studi kasus di berbagai instansi menunjukkan bahwa metode manual dapat menghambat alur kerja dan menambah waktu penyelesaian tugas administratif. Melihat tantangan tersebut, teknologi Artificial Intelligence (AI), khususnya model bahasa besar (Large Language Models - LLM) seperti Google Gemini, menawarkan solusi potensial. Dengan kemampuan pemrosesan bahasa alami (NLP) dan pemahaman multimodal (teks dan gambar), AI dapat dilatih untuk secara otomatis membaca teks dari gambar surat yang diunggah (menggunakan kapabilitas OCR implisit/eksplisit) dan mengekstraksi informasi kunci yang relevan. Implementasi AI ini diharapkan dapat mengotomatisasi proses input data, sehingga mengoptimalkan alur kerja pengarsipan digital surat masuk, meningkatkan efisiensi waktu, dan meminimalkan potensi kesalahan. Namun, penting juga untuk mempertimbangkan aspek etika dalam penggunaan AI, memastikan keamanan data dan privasi tidak dilanggar. Praktik Kerja Lapangan (PKL) ini berfokus pada penerapan AI Gemini 2.0 Flash untuk mencapai optimalisasi tersebut pada sistem pengarsipan digital di SMK DiponegoroKaranganyar.

In order to secure the usability of records over the retention period, the archives conduct digital preservation actions which may entail conversion, migration and other procedures. When using digitization for creating copies in the form of originals, the archives should consider the trustworthiness of the process - the involved professionals, equipment, software, and procedures. These are the critical points in the process of converting physical originals into digitized originals having legal validity. The authors discuss the authenticity of records converted from physical to digital form. They argue that a digitized record is a copy in the form of an original if it meets the requirements of completeness, reliability of the process, integrity and availability. The authors define 53 requirements to be met in the process of digitization.

The cultural and regional diversity across the world and specifically in India has given birth to a large number of writing systems and scripts having a variety of character sets. For scripts having a larger character set, just a simple keyboard with limited character set is not the optimal way for providing inputs to the computer. Variations in individual handwriting
due to mood swings, changes in medium of writing, changes in writing styles, etc. pose a challenge before the character recognition (CR) research community. Similar kinds of symbols in various scripts and languages act as a big barrier in multilingual CR. Lack of benchmark results and corpora for multilingual CR hinder the research in multilingual CR.
There have been only a limited number of articles for optimal combination of features and classifiers to process multilingual data. Multilingual CR has least explored the Indic scripts.
This paper presents a detailed review and analysis of the work done in multilingual online as well as offline CR for Indic and non-Indic scripts. The paper mainly contributes in two ways: Firstly, it provides a clear perspective about various phases of monolingual and multilingual CR; and secondly, identifies the major deficiencies in monolingual and multilingual CR
for printed and handwritten text. It contributes by giving an in-depth view of work done at each phase including data acquisition, pre-processing, segmentation, feature extraction, recognition and post-processing of CR. Issues to be resolved at each phase have also been
elaborated. The recent work done using Deep and Shallow architectures has been analysed. Tools used for these architectures have been compared to highlight their pros and cons. The present work also suggests how further research can be conducted in the field of monolingual
and multilingual CR. The problems such as CR in hybrid documents, identifying more reliable features, resolving issues of similar characters, identifying optimal combination strategies for deep and shallow architectures, etc. need to be tackled in future research.
Keywords Multilingual· Monolingual· Character recognition (CR)· Benchmark corpora· Indic and non-indic scripts· Japanese, Chinese, Arabic and Latin· Pre-processing·  Post-processing· Feature extraction· Deep and shallow architectures· Online CR· Offline CR

The National Library of Finland Digitisation Programme 2025–2028 reflects the goals of sustainable development. With its four-year digitisation programmes, the National Library strives to ensure the systematic selection of the material to be digitised, taking various societal groups into consideration.

The detection and extraction of scene text from document images is one of the challenging research areas. Many researchers have detected and extracted the text from plain text background. But the multi-oriented scene text detection is one of the complex problems due to multi-oriented texts which have different orientation, font size, colors etc. In this work, we have proposed a new algorithm to detect and extract the multi-oriented scene text. Experiments have been carried out to find the robustness on the proposed method, by conducting various experiments on heterogeneous datasets. The proposed method achieves selectively high detection rate of 88.43% on the multi-oriented scene text.

The code import the YoutubeTranscriptionApi from youtube_transcription_api libray, The YouTube video ID is defined. The transcription data for the given video ID is fetched using get_transcription method. 6) This methodology outlines the steps involved in converting a YouTube video transcript into text format using Python and the YouTube Transcript API. In conclusion, the provided Python script demonstrates a simple yet effective way to convert YouTube video transcripts into text format. By leveraging the youtube_transcript_api module, the script fetches the transcript data for a specified video ID and processes it to extract the text content. The resulting text is then formatted and saved to a text file for easy access and further analysis. While the current version of the script serves its basic purpose, there is ample opportunity for future enhancements and expansion. By incorporating features such as batch processing, advanced text processing, and integration with other platforms, the tool can become even more versatile and valuable to users. Overall, the project showcases the power of Python for automating tasks and working with external APIs, and it provides a solid foundation for building more sophisticated tools for video transcript processing and analysis in the future. To extract text from a video, you can use Optical Character Recognition (OCR) software or services like Adobe Acrobat, Google Drive, or online tools like OnlineOCR.net. These tools can analyze the frames of the video and convert any visible text into editable text format. If you're looking to extract specific text like project acknowledgments, you might need to manually transcribe them from the extracted text.

Optical Character Recognition (OCR) has been a transformative technology for decades, enabling the conversion of printed or handwritten text into machine-readable formats. Traditional OCR applications, built on rule-based algorithms and predefined templates, have been instrumental in digitizing structured documents such as invoices, books, and forms. However, these systems often struggle with complex layouts, poor image quality, and handwritten text due to their reliance on rigid frameworks [1]. The advent of Generative Artificial Intelligence (GenAI) has revolutionized OCR technology, addressing the limitations of traditional systems. GenAIpowered OCR leverages advanced neural networks, deep learning, and transformer-based architectures to process diverse document types with remarkable accuracy and adaptability. Unlike traditional OCR, which requires extensive manual configuration, GenAI models can dynamically learn and improve through iterative feedback loops, enabling them to handle complex layouts, varied fonts, and degraded image quality [2][3]. Traditional OCR excels in processing structured and semi-structured documents where layouts are consistent and predictable. However, its inability to adapt to new document types or interpret contextual information limits its effectiveness in modern, dynamic workflows. In contrast, GenAI-assisted OCR introduces capabilities such as contextual understanding, few-shot learning, and multimodal processing, enabling it to extract meaningful insights from unstructured and complex data sources [4][5]. This report explores the key differences between traditional OCR and GenAI-assisted OCR, highlighting their respective strengths, limitations, and applications. By examining the technological advancements and real-world use cases, this comparison aims to provide a comprehensive understanding of how GenAI is reshaping the OCR landscape and unlocking new possibilities for businesses and industries worldwide.

This article presents the findings and discussions of the “Digital Transcription of Ottoman Texts” workshop held at Marmara University in December 2023, focusing on current efforts, challenges, and innovations in the automatic transcription of Ottoman texts. The workshop brought together researchers from diverse fields to evaluate developments in handwritten text recognition (HTR), digital corpus creation, machine learning, and data augmentation methods for Ottoman Turkish, which is written in Arabic script. It highlights key projects such as LexiQamus, Osmanlica.com, and IRCICA OCR, and discusses technical hurdles like segmentation, low image resolution, and lack of standardized transcription. The article emphasizes the importance of interdisciplinary collaboration, clean and varied data sets, and the integration of large language models to improve transcription accuracy and accessibility for both scholarly and public use.
(This is a DeepL translation from Turkish to English and edited by the authors. To read the original Turkish version, click here: https://doi.org/10.18589/oa.1719330)

Diagnostic procedure, one of the most essential steps in healthcare, determines the protocol for the treatment of a patient. Doctors can provide a more accurate diagnosis and better treatment by comparing patients' recent and older test reports. However, patients typically neglect to properly store their older reports, making it more difficult to find the reports later for doctor's appointments. In this paper, we've presented an application for Analytics and Visualization of Clinical Pathology Reports or "AVCPR", which allows patients to upload and store images of their clinical pathology reports, and lets doctors access and analyze their patients' health records in an organized and more convenient than the traditional way along with dynamic visualization. Instead of relying on clinicians and healthcare organizations, AVCPR establishes a patient-centric platform where systematically stored patient records can be accessed by patients themselves, and shareable to doctors and possibly to researchers in the future. The application consists of a standard authentication system, user dashboard, profile and settings, report upload interface, automatic structured information extraction and recording system, organized record of reports with risk factor indicator color table, and comparison chart for both patients and doctors. CCS CONCEPTS • Applied Computing → Life and medical Science; Health informatics.

Tariff exemptions are fundamental to attracting Foreign Direct Investment (FDI) into the manufacturing sector, though the associated administrative processes present areas for optimization for both investing entities and the national tax authority. This paper proposes a conceptual framework to empower tax administration by leveraging a synergistic integration of Optical Character Recognition (OCR) and Large Language Model (LLM) technologies. The proposed system is designed to first utilize OCR for intelligent digitization, precisely extracting data from diverse application documents and key regulatory texts such as tariff orders. Subsequently, the LLM would enhance the capabilities of administrative officers by automating the critical and time-intensive task of verifying submitted HS Tariff Codes for machinery, equipment, and raw materials against official exemption lists. By enhancing the speed and precision of these initial assessments, this AI-driven approach systematically reduces potential for non-alignment and non-optimized exemption utilization, thereby streamlining the investment journey for FDI companies. For the national administration, the benefits include a significant boost in operational capacity, reduced administrative load, and a strengthened control environment, ultimately improving the ease of doing business and solidifying the nation's appeal as a premier destination for high-value manufacturing FDI. .

Session Title: FAIR Periodicals: What Does It Really Mean to 'Digitize' a periodical?

Panel titles:
1. FAIR Design: Metadata Strategies for Image-Driven Periodicals.
2. From OCR to AI: New Challenges and Tools in Text-Image Based Periodical Data Processing.
3. FAIR Infrastructures for the Past: Digitizing Institutional Archives and Heritage.

Periodical studies, an inherently interdisciplinary field spanning linguistics, literature, history, and cultural studies, offers a strong framework for engaging with digital humanities. Yet, as more humanities projects focus on digitization and the development of databases, digital archives, and open-access infrastructures, the challenges and opportunities of digitizing periodicals are often overlooked by scholars from non-digital humanities disciplines.
Drawing on case studies of 19th- and 20th-century Italian periodicals, this panel seeks to explore the tools and methodologies used in digitization projects, as well as the development of digital infrastructures that comply with the FAIR data principles (Findable, Accessible, Interoperable, and Reusable). Specifically, we are interested in projects that not only facilitate access to these materials but also contribute to transforming them into digital heritage.
Spanning three panels, the panel brings together diverse approaches to the digitization of periodicals. The first session focuses on institutional archives, examining both successful efforts and the practical challenges of reconstructing historical contexts through infrastructures that aim to comply with FAIR principles. The second explores visual and design-oriented periodicals, highlighting metadata strategies that not only align with FAIR standards but also grapple with the complexity of representing image-based knowledge. The final session shifts to computational tools and large-scale data analysis, investigating how technologies from OCR to AI can support—or complicate—the processing of multimodal periodical corpora. Together, these sessions offer a layered perspective on what it truly means to make periodicals Findable, Accessible, Interoperable, and Reusable in practice.

This article examines the pedagogical adaptation of Textual Communities, a digital tool originally developed for collaborative research in textual scholarship, to teach paleography at the undergraduate level within a liberal arts context. Prompted by the exigencies of remote learning during the COVID-19 pandemic and the broader framework of critical digital pedagogy, the course-design reimagined the tool's primary research-focused function-edition-makingas a dynamic teaching and learning environment emphasizing transcription, engagement, and student autonomy. The article presents a specific example of the use of a digital tool in teaching paleography, detailing its purpose and impact on student learning and engagement. The article offers a concrete case study of hybrid-and flexible-by-design pedagogy, showing the value of using scholarly digital tools in undergraduate settings.

Research Article Evaluating open-ended exams presents significant challenges in terms of time management and consistency in educational processes. This study aims to develop an iOS-based mobile application, "Exam Reader" to streamline the evaluation of handwritten open-ended exam responses by integrating visual recognition and language analysis tools, enabling educators to deliver timely and fair assessments. Developed using the Swift programming language, the application relies on two core technologies. First, handwritten student responses are converted into digital text using Optical Character Recognition (OCR) via the Google Cloud Vision API. These texts are then analyzed for clarity and coherence using the OpenAI API and GPT-4o model, ensuring that students' ideas are presented in a structured, accessible format for evaluation. Finally, the evaluation results and related data are provided to users in PDF format. Designed with a user-friendly interface, the application allows educators to quickly interpret responses and align them with expected learning outcomes through integrated language and image analysis tools. This system offers an innovative model for digitizing, standardizing, and automating open-ended exam evaluations, contributing to the systematic improvement of educational assessment processes. However, the application has limitations. Variations in handwriting and low-quality scans may reduce OCR accuracy, and AI-supported content analysis risks missing contextual nuances. Additionally, the system requires a stable internet connection, limiting offline functionality. Future enhancements, including advanced OCR models, multilingual support, and an offline mode, are planned to address these issues. The application developed in this direction is expected to make a significant contribution to the digitalization of educational assessment and to adapt to next-generation technologies.

There are two most popular writing styles of Urdu i.e. Naskh and Nastaliq. Considering Arabic OCR research, ample amount of work has been done on Naskh writing style; focusing on Urdu, which uses Arabic character set commonly used Nastaliq writing style. Due to Nastaliq writing style, Urdu OCR poses many distinct challenges like compactness, diagonal orientation and context character shape sensitivity etc., for OCR system to correctly recognize the Urdu text image. Due to compactness and slanting nature of Nastaliq writing style, existing methods for Naskh style would not give desirable results. Therefore, in this paper, we are presenting ligature based segmentation OCR system for Urdu Nastaliq script. We have discussed in detail various unique challenges for the Urdu OCR and different feature extraction techniques for Ligature recognition using SVM and kNN classifier. The system is trained to recognize 11,000 Urdu ligatures. We have achieved overall 90.29% accuracy tested on Urdu text images.

Over 8,024 wildfire incidents have been documented in
2024 alone, affecting thousands of fatalities and significant damage
to infrastructure and ecosystems. Wildfires in the United States
have inflicted devastating losses. Wildfires are becoming more
frequent and intense, which highlights how urgently efficient
warning systems are needed to avoid disastrous outcomes. The
goal of this study is to enhance the accuracy of wildfire detection
by using the Convolutional Neural Network (CNN) built on the
VGG16 architecture. The D-FIRE dataset, which includes several
kinds of wildfire and non-wildfire images, was employed in the
study. Low-resolution images, dataset imbalance, and the necessity
for real-time applicability are some of the main challenges. These
problems were resolved by enriching the dataset using data
augmentation techniques and optimizing the VGG16 model for
binary classification. The model produced a low false negative
rate, which is essential for reducing unexplored fires, despite
dataset boundaries. To help authorities execute fast responses, this
work shows that deep learning models such as VGG16 can offer a
reliable, automated approach for early wildfire recognition. To
reduce wildfire's impact, the VGG16 model achieved an accuracy
of 97.5% and produced a low false negative rate, which is crucial
for minimizing undetected fires.
Keywords—Machine Learning, Deep Learning, Wildfire Detection,
Artificial Intelligence.

There's a good deal of change in the world of insurance. The growth of the Digital Era has created open doors for technology-driven insurance claims management mechanisms. At the same time, mobile technology and AI have been used to be shared within the insurance industry. Tradition has undergone great change: one could not have imagined nor even envisioned any sort of internet-based and smartphone interfaces. The digital presence is improving the competence, transparency, and promptness of the claim's management process in insurance. Traditional backlogs-sorry for saying it: whining about paper documentation, physical verification, not keeping commitments, and complete unresponsiveness to the insured's claims-have brought trouble to the industry; particularly in the non-adhesive insurance sector, it leads to customer dissatisfaction and fineness. Customers have become increasingly impatient for those on-demand or real-time operations they can get through mobiles, which mandate insurance companies to restructure claims-filing processes so as to fit their expectations and without diverging into numerous regulatory violations or risk intents. insurers ought to implement mobile-software applications backed up by AI for faster work on claims processes, coming naturally from the drive for accomplishing greater operational efficiency, along with adopting measures to reduce the propagation of fraud and increase trust in their insureds

This paper explores the methodological and technological challenges involved in reconstructing the history of amateur filmmaking in postwar Italy through the analysis of specialized periodicals. Originating in a PhD project on Ferrania, a major Italian film stock and photographic equipment company, the research faced the absence of corporate archives and turned instead to the company’s house organ, Ferrania magazine, as a narrative and material source. Building on this experience, and within the broader scope of the PRIN 2022 project "FilmBaseMatters: A Material Approach to the History of Small-Gauge Film in Italy," the paper contributes to mapping the complex, largely undocumented landscape of Italian small-gauge cinema.

The project focuses on assembling and digitizing a corpus of magazines, newsletters, and educational journals related to small-gauge film culture. These publications — produced by cineclubs, industrial publishers, and cultural institutions — offer critical insight into the networks of amateur practice, technological dissemination, and grassroots film education. The research combines a conjectural, trace-based historical approach with digital tools for metadata tagging, OCR, and qualitative coding. However, the process has revealed substantial obstacles: bureaucratic access issues, unreliable OCR for mid-century print, and the lack of standardized controlled vocabularies, which currently limits data reusability and FAIR compliance.

Despite these limitations, the paper argues for the value of periodicals as key sources for reconstructing lost or fragmented film histories. A case study of L’altro Cinema — the Milan cineclub magazine — and its collaboration with the Museo Nazionale della Scienza e della Tecnologia Leonardo da Vinci, demonstrates how periodicals can help recover local networks, institutional alignments, and the evolving public role of amateur cinema. Ultimately, the project highlights the necessity of digital methods not only for preservation, but for enabling new analytical perspectives, making small-gauge culture visible and relevant within broader histories of cinema, media, and technology.

Optical character recognition for the English text may be considered one of the most important research topics, whether, printed or handwritten. Although excellent results have been reached in the English text, there is a lack of this type of research in the Arabic text. This is because of the nature of the Arabic alphabet, and the multiplicity of forms of the same letter. Arabic handwritten character recognition (AHCR) systems involve several issues, and challenges from finding a suitable, and public Arabic handwritten text dataset phase to recognition, and classification phase passing through segmentation, and feature extraction phases. The paper objectives are: Firstly, a large, and complex Arabic handwritten characters' dataset (HMBD) is presented for training, testing, and validation phases, as well as, discussing its collection, preparation, cleaning, and preprocessing. Secondly, we introduce a deep learning (DL) system with two convolutional neural network (CNN) architectures (named HMB1 and HMB2); with the appliance of optimization, regularization, and dropout techniques. This system can serve as a baseline for future research on handwritten Arabic text. Different performance metrics were calculated such as accuracy, recall, precision, and F1. 16 experiments were applied to the described system using HMBD, and another two datasets: CMATER, and AIA9k. Experiments' results were captured and compared to study the effects of weight initializers, optimizers, data augmentation, and regularization on overfitting, and accuracy. He Uniform weight initializer and AdaDelta optimizer reported the highest accuracies. Data augmentation showed an improvement in the accuracies. HMB1 reported testing accuracy of 98.4% with 865,840 records using augmentation on HMBD. CMATER and AIA9k datasets were used for validating the generalization. Data augmentation was applied, and the best results were 100%, and 99.0% for testing accuracies, respectively. A cross-over validation between the described architectures, and a previous state-of-the-art architecture, and dataset was performed in two phases. First, the previous control architecture cannot generalize for the presented dataset in the current study. Second, the study described architectures generalize for the control dataset, with higher accuracies (97.3%, and 96.8% for HMB1, and HMB2, respectively), than the reported accuracy in the selected control study.

The paper is a comprehensive review of the current research trends in the area of Arabic language especially state-of-the-art approaches to highlight the current status of diverse research aspects of that area to facilitate the adaption and extension of previous systems into new applications and systems. The Arabic language has deep, widespread and unexplored scope to research although the tremendous effort and researches that had been done previously. Modern state-of-the-art methods and approaches with fewer errors are required according to the high speed of hardware and technology development. The focus of this article will be on the offline Arabic handwritten text recognition as it is one of the most important topics in the Arabic scope. The main objective of this paper is critically analyzing the current researches to identify the problem areas and challenges faced by the previous researchers. This identification is intended to provide many recommendations for future advances in the area. It also compares and contrasts technical challenges, methods and the performances of handwritten text recognition previous researches works. It summarizes the critical problems and enumerates issues that should be considered when addressing these tasks. It also shows some of the Arabic datasets that can be used as inputs and benchmarks for training, testing and comparisons. Finally, it provides a fundamental comparison and discussion of some of the remaining open problems and trends in that field.

Abstract (ITA):
Per un’istituzione culturale far riferimento all’ecosistema IIIF non significa certo uniformarsi a una moda tecnologica del momento, ma molto più opportunamente rappresenta a tutt’oggi la via migliore per assicurare ai contenuti digitalizzati che si intende diffondere la possibilità di divenire pienamente usabili e ricercabili, nel pieno rispetto delle buone pratiche dei LOD-Linked Open Data su cui si fonda il Semantic Web.
L’ampia adozione dell’architettura IIIF ha cambiato radicalmente il paradigma di utilizzo degli oggetti digitali. Prima della diffusione di tale tecnologia, infatti, molti progetti di digitalizzazione che le biblioteche e le istituzioni culturali in generale stavano implementando, fornivano sì l’accesso a collezioni preziose, ma esse venivano considerate e gestite come dei silos isolati. L’accesso a queste ricche collezioni di oggetti digitalizzati dipendeva in modo cruciale dalla configurazione di ciascun database e dalle funzioni del singolo servizio: una vera complicazione per gli utenti online, che si trovavano costretti a dover ripetere più volte la propria ricerca online per la visualizzazione dei contenuti digitalizzati presenti all’interno di repositories diversi.
Un mezzo per abbattere queste barriere è stato proprio l’impiego delle tecniche del Semantic Web, in grado di interconnettere i depositi di oggetti digitali altrimenti isolati e autoreferenziali. In sostanza, il modello di dati IIIF si basa sull’approccio del Semantic Web per consentire agli studiosi di confrontare e studiare gli oggetti digitali utilizzando strumenti in grado di offrire una visualizzazione simultanea delle risorse nonostante le loro diverse provenienze. Il termine “interoperabilità” contenuto nell’acronimo IIIF indica proprio tale sforzo tecnologico che cerca di soddisfare questa esigenza per esprimere l’opportunità di scambiare informazioni o servizi tra sistemi informatici diversi, facilitando la loro reciproca interazione.

- - -

Abstract (ENG):
The mention of the IIIF ecosystem by cultural institutions is not just a passing trend but represents a crucial method to ensure that digitized content is fully accessible and searchable according to best practices of LOD-Linked Open Data and the Semantic Web. The adoption of IIIF architecture has transformed how digital objects are used, moving away from isolated silos to interconnected repositories that enhance access to valuable collections. Previously, digitization projects led to fragmented access based on database configurations, complicating user experiences. Semantic Web techniques have helped break down these barriers by connecting disparate digital repositories, improving the accessibility and usability of digitized content found in different repositories. Overall, the IIIF data model leverages Semantic Web principles to facilitate seamless access and search functionalities for digitized materials.

Dieser Beitrag stellt eine neuartige Methode zur optischen Zeichenerkennung ( Optical Character Recognition, OCR) speziell für Textvorlagen des 17. Jahrhunderts vor. Anstatt ein neues OCR-Verfahren zu entwickeln, werden zwei etablierte Open-Source-Lösungen genutzt. Die Ausgaben der Programme werden computergestützt kombiniert, um so eine möglichst genaues Textergebnis zu erhalten. Die Besonderheiten und die Güte der Methode wird anhand der Texterfassung von Gelegenheitsgedichten von Simon Dach illustriert.

The handwriting recognition is the scheme of converting text symbolized in the spatial form of graphical symbols into its figurative depiction. Handwritten characters have been the most accredited technique of collecting, storing and transmitting information all the way through the centuries. To give the proper ability to the machine it requires studying the image-form of data which forms a special pattern to be interpreted. Designing and building machines that can recognize patterns remains one of the thrust areas in the field of computer sciences. A lot of work has been done in this field, but still the problem is not answered in its full density. A good text recognizer has many commercial and practical applications, e.g. from finding data in digitized book to computerization of any organization, like post office, which involve manual task of interpreting text. In this paper, we have presented a hybrid approach for recognition of Gujarati handwritten numerals using neural networks as classifier and achieved a good recognition rate for noisy numerals.

Between the 16th and 18th centuries, Spanish and Indigenous people produced millions of colonial documents in diverse calligraphic styles in Latin American territories. Accessing their rich information requires specialised palaeographic skills and often indigenous language expertise. Recent advancements in Machine Learning offer promising solutions to these challenges. This study developed two computational methods: (1) a historical document classifier using Convolutional Neural Networks (CNNs) and Support Vector Machines (SVMs), and (2) Handwritten Text Recognition (HTR) models trained on 16th- and 17th-century Spanish manuscripts via Transkribus. The classifier achieved high accuracy, with F1 scores over 90% for most calligraphic styles, while HTR models produced competitive Character Error Rates, notably 5.25% for Redonda, 8.92% for Itálica Cursiva and 14.15% Procesal Simple. These automated tools are now allowing us to transform previously “unreadable” archives into accessible data. This means that soon, Latin American archives and libraries will be able to unlock centuries of historical information.

Contemporary healthcare techniques
involve maintaining a physical record of
doctor visits, prescriptions, medical
findings, tests, etc. Receiving our medical
reports on WhatsApp, calling our doctor, or
searching for medicines online is the closest
we have come to digitization of healthcare.
Coupled down by the fact that a lot of
individuals know little to know about their
medical history, prescription makes it
challenging to treat them in critical
conditions. Fragmentation of medical
records, leading to misdiagnoses, redundant
tests, and treatment inefficiencies.

Cuneiform language is an old language that was invented by the people of Sumerian nation. It is an essential language for many archeologists. Especially who are interested in studying and investigating the old nations of Iraq. Dealing with this type of language usually requires specialist to translate its symbols, which are basically forms of nail shapes. This study presents a new approach to translate the cuneiform writing by employing artificial neural network (ANN) technique. Effectively, multi-layer perceptron (MLP) neural network has been adapted for translating the Sumerian cuneiform symbol images to their corresponding English letters. This work has been successfully established and it attained 100%.

The paper demonstrates that malware can be detected through supervised machine learning. Using machine learning for malware detection has technical and commercial benefits. This is because malware detection in the current context is about constantly researching for new signatures developed by hackers and programming for detecting these signatures in files. This process is effort intensive and time intensive. The result is an expensive process for malware detection. If the machine could be made to detect malware, not only will the process of malware detection become less expensive due to less human effort involved; the time required to detect new malware will reduce drastically.

The form processing systems commercially available include a verification step during which a human operator verifies the output provided by the system to ensure 100% accuracy. In order to reduce the time and the cost of such a stage, the OCR engine incorporated into the system provides a reliability measure of the classification to be used for implementing a reject option: in this way only rejected samples are passed to the verification stage. Most of the strategies for designing such a reject option consider that the source of classification errors are within the OCR engine. Such an assumption becomes less reasonable as the forms become less structured, as in case when boxes are provided for the entire data field and not only for isolated characters. Under these circumstances, we investigate to which extent the reliability measure provided by an OCR engine designed to deal with boxed isolated characters can be used to detect both segmentation and classification errors. The experimental results, obtained on a large data set of forms currently in use by a large organization, show that the proposed method successfully achieves its aim. It represents a powerful tool for the system manager to plan system enhancement as the volume of forms containing less constrained data fields increases.

The admission documentation process for Indian students is often time-consuming and labourintensive. This research explores leveraging Optical Character Recognition (OCR), generative Artificial Intelligence (AI), and automation to streamline this process. By developing a web application that integrates these technologies, we propose a system where students can upload documents, validated for authenticity using generative AI, scanned using OCR for data extraction, and automatically used to fill admission forms. The proposed system can reduce the time and effort involved for both students and administrative staff, improve efficiency, reduce errors, and provide a scalable solution for various documentation needs. This theoretical research highlights the potential benefits and lays the groundwork for future development and real-world applications, including government and healthcare sectors.

Handwriting character recognition is an active and challenging area of research in the pattern-recognition field. Nowadays, many algorithms are introduced to achieve the highest accuracy. An improved recognition result could be accomplished if the input character images have good quality. That is why the preprocessing step becomes crucial for image identification missions. The algorithm of CNN (Convolutional Neural Networks) is fortified with numerous designs, permitting researchers to select the most operative architecture for better classification. However, this study suggests that the pre-processing mechanism is an important factor to be considered to increase the identification accuracy. This study utilised seven stages for pre-processing mechanisms applied to the dataset then the classification-accuracy measurement resultant from CNN was acquired. This work exposed that the utilisation of pre-processing steps indicated the most heightened accuracy with Binarized-KDS as an input, achieving training, testing, and validation accuracy of 99.2%, 97%, and 97.2% respectively after 35 iterations. Furthermore, another significant finding is that using different steps in processing input images to train the recognition model affects the recognition-rate within the same classifiers. Besides, the outcome reveals that each technique applied with the specific classifier may require a certain pre-process to obtain its optimal accuracy recognition rate.

In this presentation, we briefly describe five projects of analysis and interpretation of the mortuary texts written in Earlier Egyptian, that is, ancient and middle Egyptian (between 2500 and 1500 BC). Most of the texts under study are inscribed inside rectangular coffins, of which only roughly two hundred are published minimally adequately. The texts consist of copies of the famous Pyramid Texts and the no less famous Coffin Texts. These two terms are Egyptological, not Egyptian, and assume that there is a chronological sequence between both corpora; a point that needs to be reviewed. The use of digital technology can contribute to this aspect, but the lines of research implemented through these five projects have two fundamental, much more practical objectives:

    Accelerate the research community's access to the largest corpus of texts ever written in ancient Egypt with a powerful tool that will allow for in-depth, large-scale analysis of graphemics, morphology, syntax and semantics, as well as interdisciplinary studies in history, religion, anthropology and archaeology.
    Provide universal access to these texts through the Internet, following the basic principle of open science.

https://www.transkribus.org/model/agapet-13th-century As part of the Agapet project, which aims to train with Transkribus HTR models to digitize Christian Arabic literature from the 9th to the 19th century, we present a new model, Agapet13, for the manuscripts of the 13th century. This development follows the training of Agapet17 for the manuscripts of the 17th century: https://www.transkribus.org/model/agapet17-christian-arabicmanuscripts. Agapet13 (Transkribus Model ID 300833) is an advanced Handwritten Text Recognition (HTR) model designed to transcribe Christian Arabic manuscripts, with a specific focus on the script used by the renowned 13th-century Melkite scribe, Poimen al-Sīqī. The model, developed to facilitate the study and accessibility of Christian Arabic textual heritage, is trained on a dataset of 307 pages comprising approximately 59,103 words. It currently achieves a Character Error Rate (CER) of 7.48%, making it a promising tool for scholars working on Arabic manuscript studies.

Persons of visual impairment make up a growing segment of modern society. To cater to the special needs of these individuals, society ought to consider the design of special constructs to enable them to fulfill their daily necessities. This research proposes a new method for text extraction from indoor signage that will help persons of visual impairment maneuver in unfamiliar indoor environments, thus enhancing their independence and quality of life. In this thesis, images are acquired through a video camera mounted on glasses of the walking person. Frames are then extracted and used in an integrated framework that applies Maximally Stable Extremal Regions (MSER) to detect alphabets along with a morphological dilation operation to identify clusters of alphabets (words). Proposed method has the ability to localize and detect the orientation of these clusters. A rotation transformation is performed when needed to realign the text into a horizontal orientation and allow the objects to be in an acceptable input to any of the available optical character recognition (OCR) systems. Analytical and simulation results verify the validity of the proposed system.

Newspapers are among the most comprehensive textual records from late Ottoman Palestine that were produced by the inhabitants of the region themselves. When subjected to ‘close reading,’ they can provide additional information about individual people and places named in Ottoman census registers.

This research focuses on real-time surveillance systems as a means for tackling the issue of non-compliance with helmet regulations, a practice that considerably amplifies the risk for motorcycle drivers or riders. Despite the well-established advantages of helmet usage, achieving widespread compliance remains challenging due to diverse contributing factors. To effectively address this concern, real-time monitoring and enforcement of helmet laws have been proposed as a plausible solution. However, previous attempts at real-time helmet violation detection have been hindered by their limited ability to operate in real-time. To overcome this limitation, the current paper introduces a novel real-time helmet violation detection system that utilizes the YOLOv5 single-stage object detection model. This model is trained on the 2023 NVIDIA AI City Challenge 2023 Track 5 dataset, which includes 100 videos used for training and testing, each with an average duration of 20 seconds. The optimal hyperparameters for training the model are determined using genetic algorithms. Additionally, data augmentation and various sampling techniques are implemented to enhance the model's performance. The efficacy of the models is evaluated using precision, recall, and mean Average Precision (mAP) metrics. The results demonstrate impressive precision, recall, and mAP scores of 0.848, 0.599, and 0.641, respectively for the training data. Furthermore, the model achieves notable mAP score of 0.6667 for the test datasets, leading to a commendable 4 th place rank in the public leaderboard. This innovative approach represents a notable breakthrough in the field and holds immense potential to substantially enhance motorcycle safety. By enabling real-time monitoring and enforcement capabilities, this system has the capacity to contribute towards increased compliance with helmet laws, thereby effectively reducing the risks faced by motorcycle riders and passengers.

Cuneiform documents, the earliest known form of writing, are prolific textual sources of the ancient past. Experts publish editions of these texts in transliteration using specialized typesetting, but most remain inaccessible for computational analysis in traditional printed books or legacy materials. Off-the-shelf OCR systems are insufficient for digitization without adaptation. We present CuReD (Cuneiform Recognition-Documents), a deep learning-based human-in-the-loop OCR pipeline for digitizing scanned transliterations of cuneiform texts. CuReD has a character error rate of 9% on clean data and 11% on representative scans. We digitized a challenging sample of transliterated cuneiform documents, as well as lexical index cards from the University of Pennsylvania Museum, demonstrating the feasibility of our platform for enabling computational analysis and bolstering machine-readable cuneiform text datasets. Our result provide the first human-in-the-loop pipeline and interface for digitizing transliterated cuneiform sources and legacy materials, enabling the enrichment of digital sources of these low-resource languages.

At the Faculty of Humanities and Social Sciences, University of Zagreb, the Transcribathon Zagreb event took place from November 7 to 11, 2022. This event combined a scientific-professional conference and a student competition in transcribing archival materials. The main organizer was the State Archives in Zagreb, which is also a partner in the Enrich Europeana+ project (*Enriching Europeana through citizen science and artificial intelligence – unlocking the 19th century*). The conference segment, titled *"Cultural Heritage and AI: Transcribathon Zagreb,"* lasted two days, from November 7 to 8, 2022, featuring a total of 13 presentations by domestic and international experts in the field of archival science. International experts included Ad Polle (Europeana), who delivered the keynote presentation, and Bojan Cvelfar (Archives of the Republic of Slovenia), who presented *"The e-ARH.si Project and Its Key Solutions."* As part of the conference, a guided tour of the city of Zagreb was also organized.

This project focuses on developing a personalized allergen notification system to help individuals identify potential allergens in food products based on their dietary preferences and known allergens. A key feature of this system is Optical Character Recognition (OCR), which enables users to scan ingredient lists directly from product packaging. The OCR technology extracts text from images of ingredients, which is then processed using natural language processing (NLP) and machine learning algorithms to compare the extracted text against the user's allergen profile. The system provides real-time notifications of potential allergens, offering an efficient and portable solution for allergen detection and food safety.

utomatic transcription technology, or handwritten text recognition (HTR), unlocks new opportunities for analysing historical texts by transforming document images into machine-readable formats. This paper explores the development and evaluation of an image pre-processing pipeline to improve transcription accuracy for Tibetan historical newspapers. Images of historical texts often suffer from degradations introduced throughout their lifecycle, negatively affecting HTR accuracy. Additionally, the variability in image quality across archives poses challenges to creating a universally applicable pre-processing pipeline. This study compares three pre-processing pipelines, ultimately revealing a dynamic approach that could adapt to varying document conditions and that resulted in the highest transcription accuracy. This method offers a replicable solution for future research. We have also made the source code publicly available to support further exploration.

The retail industry, especially grocery and liquor store, faces unprecedented challenges due to a 1.2 million worker shortage in the U.S., leading to deteriorating service quality, inadequate inventory management, and alarming levels of theft. Shoplifting alone has surged by 60% in the past year, resulting in annual losses of $69 billion for retailers, and store closures have increased by 30%. Autonomous stores have emerged as a transformative solution, eliminating customer waiting times and enabling 24/7 operation. However, their success relies heavily on advancements in product recognition technologies. This research focuses on addressing the complex challenge of recognizing grocery and liquor products, which is compounded by the vast variety and subtle differences in product appearance. This research addresses critical challenges in retail product recognition through the development of a novel multimodal approach combining computer vision and natural language processing. We introduce the Liquor-grocery dataset, featuring high-resolution 360-degree product images and associated textual data from 538 distinct classes. Our methodology leverages state-of-the-art object detection models, particularly YOLO variants, optimized through extensive hyperparameter tuning. By integrating visual and textual features, we demonstrate significant improvements in fine-grained product classification accuracy and efficiency compared to existing approaches. The proposed system achieves over 93% mean average precision on our dataset, with real-time inference capabilities suitable for practical retail applications. This work contributes to advancing automated retail technologies, offering potential enhancements in inventory management, self-checkout systems, and customer experience. The successful adoption of these technologies has the potential to revolutionize the industry, enabling stores to operate 24/7, reduce losses from shoplifting, and optimize inventory management, contributing to a more resilient and sustainable retail landscape.