Optical Character Recognition Output Research Articles

High Throughput Information Extraction of Printed Specimen Labels from Large-Scale Digitization of Entomological Collections using a Semi-Automated Pipeline

Insects account for half of the total described living organisms on Earth, with a vast number of species awaiting description. Insects play a major role in ecosystems but are yet threatened by habitat destruction, intensive farming, and climate change. Museum collections around the world house millions of insect specimens and large-scale digitization initiatives, such as the digitization street digitize! at the Museum für Naturkunde, have been undertaken recently to unlock this data. Accurate and efficient extraction of insect specimen label information is vital for building comprehensive databases and facilitating scientific investigations, sustainability of the collected data, and efficient knowledge transfer. Despite the advancements in high-throughput imaging techniques for specimens and their labels, the process of transcribing label information remains mostly manual and lags behind the pace of digitization efforts. In order to address this issue, we propose a three step semi-automated pipeline that focuses on extracting and processing information from individual insect labels. Our solution is primarily designed for printed insect labels, as the OCR (optical character recognition) technology performs well for printed text while handwritten texts still yield mixed results. The pipeline incorporates computer vision (CV) techniques, OCR, and a clustering algorithm. The initial stage of our pipeline involves image analysis using a convolutional neural network (CNN) model. The model was trained using 2100 images from three distinct insect label datasets, namely AntWeb (ant specimen labels from various collections), Bees & Bytes (bee specimen labels from the Museum für Naturkunde), and LEP_PHIL (Lepidoptera specimen labels from the Museum für Naturkunde). The first model enables the identification and isolation of single labels within an image, effectively segmenting the label region from the rest of the image, and crops them into multiple new, single-label image files. It also assigns the labels to different classes, i.e., printed text or handwritten, with handwritten labels sorted out from the printed ones. In the second step, labels classified as “printed” are then parsed by an OCR engine to extract the text information from the labels. Tesseract and Google Vision OCRs were both tested to assess their performance. While Google Vision OCR is a cloud-based service with limited configurability, Tesseract provides the flexibility to fine-tune settings and enhance its performance for our specific use cases. In the third step, the OCR outputs are aggregated by similarity using a clustering algorithm. This step allows for the identification and formation of clusters that consist of labels sharing identical or highly similar content. Ultimately, these clusters are compared against a curated database of labels and are assigned to a known label or highlighted as new and manually added to the database. In order to assess the efficiency of our pipeline, we performed benchmarking experiments using a set of images similar to those the models were trained on, as well as additional image sets obtained from various museum collections. Our pipeline offers several advantages, streamlining the data entry process, and reducing manual extraction time and effort, while also minimizing potential human errors and inconsistencies in label transcription. The pipeline holds the promise of accelerating metadata extraction from insect specimens, promoting scientific research and enabling large-scale analyses to achieve a more profound understanding of the collections.

IMPROVING THE IMAGE-TO-SPEECH SYSTEM ACCURACY THROUGH INTEGRATION OF OPTICAL CHARACTER RECOGNITION AND LANGUAGE PROCESSING TECHNIQUES

Image-to-speech systems are a type of technology allowing for the conversion of visual information, such as images or videos, into auditory output. These systems use complex algorithms and machine learning techniques to recognize and describe visual content, allowing individuals who are visually impaired or blind to access in-formation that would otherwise be inaccessible to them. Image-to-speech systems are becoming increasingly sophisticated and can be integrated into a variety of devices, from smartphones to smart glasses. This article presents an approach to improving the accuracy of the image-to-speech system by incorporating multiple techniques. The proposed system begins by using Tesseract, an optical character recognition (OCR) engine, to extract text infor-mation from images. However, OCR is often imperfect and produces errors, which can impact the accuracy of image-to-speech models. To address this issue, the Text-Davinci-002 engine was applied for post-processing OCR output, which can help to correct errors and improve the accuracy of the extracted text. Finally, the Microsoft Speech API was employed in order to generate speech from the extracted text. By integrating these three techniques, image-to-speech system accuracy was significantly improved. An example of the generated synthetic dataset showed that the proposed techniques improve image-to-speech system accuracy both on word and character levels, and also perform punctuation error correction. This approach can be useful in various applications, including reading text from images, translating written text to speech, and assisting people with visual im-pairments.

Overall Equipment Utilisation (OEU) Monitoring and Remote Quality Check in Legacy Machine with Raspberry Pi

Overall Equipment Utilisation (OEU) plays an important role as a benchmark for manufacturing companies to determine each machine's efficiency. Currently, there is no proper OEU measurement system in legacy machines and only relies on human observation. This project aims to develop a measurement of OEU system by using Optical Character Recognition (OCR). An efficient Optical Character Recognition (OCR) algorithm is needed to have a high percentage of recognition rate. The outcome of this project will be a Graphical User Interface (GUI) that display real-time OEU monitoring and remote quality check for legacy machines. Pytesseract-OCR Version 4 classifier using the Recurrent Neural Network (RNN) method has been proposed in this paper. Furthermore, an error detection feature is designed from OCR output.

OCR error correction using correction patterns and self-organizing migrating algorithm

Optical character recognition (OCR) systems help to digitize paper-based historical achieves. However, poor quality of scanned documents and limitations of text recognition techniques result in different kinds of errors in OCR outputs. Post-processing is an essential step in improving the output quality of OCR systems by detecting and cleaning the errors. In this paper, we present an automatic model consisting of both error detection and error correction phases for OCR post-processing. We propose a novel approach of OCR post-processing error correction using correction pattern edits and evolutionary algorithm which has been mainly used for solving optimization problems. Our model adopts a variant of the self-organizing migrating algorithm along with a fitness function based on modifications of important linguistic features. We illustrate how to construct the table of correction pattern edits involving all types of edit operations and being directly learned from the training dataset. Through efficient settings of the algorithm parameters, our model can be performed with high-quality candidate generation and error correction. The experimental results show that our proposed approach outperforms various baseline approaches as evaluated on the benchmark dataset of ICDAR 2017 Post-OCR text correction competition.

Business Process Automation: A Workflow Incorporating Optical Character Recognition and Approximate String and Pattern Matching for Solving Practical Industry Problems

Companies are relying more on artificial intelligence and machine learning in order to enhance and automate existing business processes. While the power of OCR (Optical Character Recognition) technologies can be harnessed for the digitization of image data, the digitalized text still needs to be validated and enhanced to ensure that data quality standards are met for the data to be usable. This research paper focuses on finding and creating an automated workflow that can follow image digitization and produce a dictionary consisting of the desired information. The workflow introduced consists of a three-step process that is implemented after the OCR output has been generated. With the introduction of each step, the accuracy of key-value matches of field names and values is increased. The first step takes the raw OCR output and identifies field names using exact string matching and field-values using regular expressions from an externally maintained file. The second step introduces index pairing that matches field-values to field names based on the location of the field name and value on the document. Finally, approximate string matching is introduced to the workflow, which increases accuracy. By implementing these steps, the F-measure for key-value pair matches is measured at 60.18% in the first step, 80.61% once index pairing is introduced, and finally 90.06% after approximate string matching is introduced. The research proved that accurate usable data can be obtained automatically from images with the implementation of a workflow after OCR.

Camlens – an Innovative Android Phone Application to Empower the Blind and Visually Impaired in Reading any Kind of Printed Text in Real-Time using Opencv, Optical Character Recognition and Text-To-Speech

CamLens is an innovative mobile phone application with the vision to empower the blind and visually impaired people in reading any kind of printed text. CamLens intelligently scans the frame inside the camera view to automatically select the exact edges of the document containing the text in real-time; crop the file from the four selected edges and alternate the perspective transformation of the cropped photographto obtain an image containing the scanned document. The application then applies Optical Character Recognition (OCR) and (TTS) conversion to extract the text from the image and to convert the extracted text into audio output respectively which may be listened by the blind and visually impaired person the usage of earphones.The genesis of the research comes from the fact that the three edges of a page of the book are easier to find with lesser possibilities of finding the fourth edge precisely which, in turn, hampers the cropping accuracy as well as the OCR and TTS output.

Mass Digitization of Early Modern Texts With Optical Character Recognition

Optical character recognition (OCR) engines work poorly on texts published with premodern printing technologies. Engaging the key technological contributors from the IMPACT project, an earlier project attempting to solve the OCR problem for early modern and modern texts, the Early Modern OCR Project (eMOP) of Texas A8M received funding from the Andrew W. Mellon Foundation to improve OCR outputs for early modern texts from the Eighteenth Century Collections Online (ECCO) and Early English Books Online (EEBO) proprietary database products—or some 45 million pages. Added to print problems are the poor quality of the page images in these collections, which would be too time consuming and expensive to reimage. This article describes eMOP's attempts to OCR 307,000 documents digitized from microfilm to make our cultural heritage available for current and future researchers. We describe the reasoning behind our choices as we undertook the project based on other relevant studies; discoveries we made; the data and the system we developed for processing it; the software, algorithms, training procedures, and tools that we developed; and future directions that should be taken for further work in developing OCR engines for cultural heritage materials.

Towards Mobile OCR: How To Take a Good Picture of a Document Without Sight.

The advent of mobile OCR (optical character recognition) applications on regular smartphones holds great promise for enabling blind people to access printed information. Unfortunately, these systems suffer from a problem: in order for OCR output to be meaningful, a well-framed image of the document needs to be taken, something that is difficult to do without sight. This contribution presents an experimental investigation of how blind people position and orient a camera phone while acquiring document images. We developed experimental software to investigate if verbal guidance aids in the acquisition of OCR-readable images without sight. We report on our participant's feedback and performance before and after assistance from our software.

Integrating natural language processing with image document analysis: what we learned from two real-world applications

Automatically accessing information from unconstrained image documents has important applications in business and government operations. These real-world applications typically combine optical character recognition (OCR) with language and information technologies, such as machine translation (MT) and keyword spotting. OCR output has errors and presents unique challenges to late-stage processing. This paper addresses two of these challenges: (1) translating the output from Arabic handwriting OCR which lacks reliable sentence boundary markers, and (2) searching named entities which do not exist in the OCR vocabulary, therefore, completely missing from Arabic handwriting OCR output. We address these challenges by leveraging natural language processing technologies, specifically conditional random field-based sentence boundary detection and out-of-vocabulary (OOV) name detection. This approach significantly improves our state-of-the-art MT system and achieves MT scores close to that achieved by human segmentation. The output from OOV name detection was used as a novel feature for discriminative reranking, which significantly reduced the false alarm rate of OOV name search on OCR output. Our experiments also show substantial performance gains from integrating a variety of features from multiple resources, such as linguistic analysis, image layout analysis, and image text recognition.

Natural Language Processing for Historical Texts Michael Piotrowski (Leibniz Institute of European History) Morgan & Claypool (Synthesis Lectures on Human Language Technologies, edited by Graeme Hirst, volume 17), 2012, ix+157 pp; paperbound, ISBN 978-1608459469

<b>Natural Language Processing for Historical Texts</b> <b>Michael Piotrowski</b> (Leibniz Institute of European History) Morgan &amp; Claypool (Synthesis Lectures on Human Language Technologies, edited by Graeme Hirst, volume 17), 2012, ix+157 pp; paperbound, ISBN 978-1608459469

Book History in the Early Modern OCR Project, or, Bringing Balance to the Force

The Early Modern OCR Project (eMOP), funded by a development grant from the Andrew Mellon Foundation to Texas AM instead, they find their power source in that bibliograph- ical data. In the case of eMOP, as this essay will discuss, the relationship between the digital and the bibliographic is dialogic and reciprocal: while the project's goals, angles of approach, and ethos of interdisciplinarity are all char- acteristic of DH, it is only through an acknowledged utilization of book his- tory scholarship and methods that the project's ends will be accomplished. Book history-our corner of eMOP-represents two foundational nodes of the project. In the first place, we are identifying specific, minutely variant typefaces in order to distinguish as best we can between the myriad versions of the standard Roman typeface in early modern books.1 Secondly, we are study- ing type founders and foundries to trace the flow of fonts into and through London. Through this research we hope to realize the goal of eMOP: the auto- mation of a process by which trained optical character recognition (OCR) en- gines might more accurately the images of early modern book pages in, for example, Early English Books Online (EEBO) and Eighteenth Century Collec- tions Online (ECCO). Ultimately our work will be formalized in a database that serves as the hub of this automated OCR process: the printer and typo- graphical data will act as a traffic cop of sorts, directing the properly trained OCR engine to read the appropriate page images. In fact, a central tenet of the Early Modern OCR Project is that training OCR engines to recognize the let- terforms in specific font sets will improve the accuracy of the OCR output- the resultant text files-when these engines are called upon to scan page im- ages printed in that typeface.While this is only a cursory sketch of eMOP, it suggests that book history is central to solving the digital problem of using OCR software on early mod- ern books. Before printing became more regularized by technological ad- vances in the nineteenth century, and before English typography approached something of a standard, national identity in the early eighteenth century, the typefaces and their settings on the printed page were highly variable. For this reason one might think that the most advanced OCR engines of our day might be rather effective, as they pull from their expansive font libraries to recognize different characters. However, part of the complication is that a large number of characters found in early English printing-including the ubiquitous long s, scribal abbreviations borrowed from the Latin, and even characters derived from Anglo-Saxon, such as the thorn-are unfamiliar to non-expert readers and not present in OCR libraries. Another challenge lies in the range of type- faces used by printers, which were for most stretches of English printing im- ported from various other countries and so display characteristics adopted from these national traditions. It is still another kind of noise, however, that makes them ineffective: because the engines cannot distinguish line divi- sions-cannot focus their field of vision, in other words-they are not able to discriminate between letterforms and the various other blots that cloud page images. On more modern, higher quality images, of course, the OCR is more accurate, but those that have been preserved in mass-digitization projects- and which therefore will be central to eMOP's automated process-were lim- ited by the technologies of their historical moments. …

Unsupervised profiling of OCRed historical documents

In search engines and digital libraries, more and more OCRed historical documents become available. Still, access to these texts is often not satisfactory due to two problems: first, the quality of optical character recognition (OCR) on historical texts is often surprisingly low; second, historical spelling variation represents a barrier for search even if texts are properly reconstructed. As one step towards a solution we introduce a method that automatically computes a two-channel profile from an OCRed historical text. The profile includes (1) “global” information on typical recognition errors found in the OCR output, typical patterns for historical spelling variation, vocabulary and word frequencies in the underlying text, and (2) “local” hypotheses on OCR-errors and historical orthography of particular tokens of the OCR output. We argue that availability of this kind of knowledge represents a key step for improving OCR and Information Retrieval (IR) on historical texts: profiles can be used, e.g., to automatically finetune postcorrection systems or adapt OCR engines to the given input document, and to define refined models for approximate search that are aware of the kind of language variation found in a specific document. Our evaluation results show a strong correlation between the true distribution of spelling variation patterns and recognition errors in the OCRed text and estimated ranks and scores automatically computed in profiles. As a specific application we show how to improve the output of a commercial OCR engine using profiles in a postcorrection system.

Robust named entity detection from optical character recognition output

In this paper, we focus on information extraction from optical character recognition (OCR) output. Since the content from OCR inherently has many errors, we present robust algorithms for information extraction from OCR lattices instead of merely looking them up in the top-choice (1-best) OCR output. Specifically, we address the challenge of named entity detection in noisy OCR output and show that searching for named entities in the recognition lattice significantly improves detection accuracy over 1-best search. While lattice-based named entity (NE) detection improves NE recall from OCR output, there are two problems with this approach: (1) the number of false alarms can be prohibitive for certain applications and (2) lattice-based search is computationally more expensive than 1-best NE lookup. To mitigate the above challenges, we present techniques for reducing false alarms using confidence measures and for reducing the amount of computation involved in performing the NE search. Furthermore, to demonstrate that our techniques are applicable across multiple domains and languages, we experiment with optical character recognition systems for videotext in English and scanned handwritten text in Arabic.

Automated system for text detection in individual video images

Text detection in video images is a challenging research problem because of the poor spatial resolution and complex back- ground, which may contain a variety of colors. An automated system for text detection in video images is presented. It makes use of four modules to implement a series of processes to extract text regions from video images. The first module, called the multistage pulse code modulation (MPCM) module, is used to locate potential text regions in color video images. It converts a video image to a coded image, with each pixel encoded by a priority code ranging from 7 down to 0 in accordance with its priority, and further produces a binary thresholded image, which segments potential text regions from the background. The second module, called the text region detection module, applies a sequence of spatial filters to remove noisy regions and eliminate regions that are unlikely to contain text. The third module, called the text box finding module, merges text regions and produces boxes that are likely to contain text. Finally, the fourth module, called the optical character recognition (OCR) module, eliminates the text boxes that produce no OCR output. An extensive set of experiments is conducted and demonstrates that the proposed system is effective in detecting text in a wide variety of video images. © 2003 SPIE and IS&T. (DOI: 10.1117/1.1584050)