Attention-based Fully Gated CNN-BGRU for Russian Handwritten Text

Abdallah, Abdelrahman; Hamada, Mohamed; Nurseitov, Daniyar

doi:10.48550/arxiv.2008.05373

Cited by 3 publications

(2 citation statements)

References 29 publications

(35 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In the case of Eastern European languages, the Kazakh Offline Handwritten Text Dataset (KOHTD) (Toiganbayeva et al ., 2022), a dataset of Cyrillic exam papers, and the Handwritten Kazakh and Russian (HKR) database, have recently emerged (Nurseitov et al ., 2021). These databases are essential for training and the eventual recognition of texts, allowing these materials to be consulted and analysed for the first time, as exemplified by Abdallah et al . (2020, p. 141) producing models with CERs as low as 0.045% and WERs as low as 0.192% using the HKR dataset.…”

Section: Resultsmentioning

confidence: 99%

The implications of handwritten text recognition for accessing the past at scale

Nockels,

Gooding,

Terras

2024

View full text Add to dashboard Cite

PurposeThis paper focuses on image-to-text manuscript processing through Handwritten Text Recognition (HTR), a Machine Learning (ML) approach enabled by Artificial Intelligence (AI). With HTR now achieving high levels of accuracy, we consider its potential impact on our near-future information environment and knowledge of the past.Design/methodology/approachIn undertaking a more constructivist analysis, we identified gaps in the current literature through a Grounded Theory Method (GTM). This guided an iterative process of concept mapping through writing sprints in workshop settings. We identified, explored and confirmed themes through group discussion and a further interrogation of relevant literature, until reaching saturation.FindingsCatalogued as part of our GTM, 120 published texts underpin this paper. We found that HTR facilitates accurate transcription and dataset cleaning, while facilitating access to a variety of historical material. HTR contributes to a virtuous cycle of dataset production and can inform the development of online cataloguing. However, current limitations include dependency on digitisation pipelines, potential archival history omission and entrenchment of bias. We also cite near-future HTR considerations. These include encouraging open access, integrating advanced AI processes and metadata extraction; legal and moral issues surrounding copyright and data ethics; crediting individuals’ transcription contributions and HTR’s environmental costs.Originality/valueOur research produces a set of best practice recommendations for researchers, data providers and memory institutions, surrounding HTR use. This forms an initial, though not comprehensive, blueprint for directing future HTR research. In pursuing this, the narrative that HTR’s speed and efficiency will simply transform scholarship in archives is deconstructed.

show abstract

Section: Resultsmentioning

confidence: 99%

The implications of handwritten text recognition for accessing the past at scale

Nockels,

Gooding,

Terras

2024

View full text Add to dashboard Cite

show abstract

“…A new attention-based fully gated convolutional RNN was proposed by Abdallah [24], this model was trained and tested on the HKR dataset [4]. This work shows the effect of the attention mechanism and the gated layer on selecting relative features.…”

Section: Related Workmentioning

confidence: 99%

Classification of Handwritten Names of Cities and Handwritten Text Recognition using Various Deep Learning Models

Nurseitov¹,

Bostanbekov²,

Kanatov³

et al. 2020

Adv. sci. technol. eng. syst. j.

Self Cite

View full text Add to dashboard Cite

This article discusses the problem of handwriting recognition in Kazakh and Russian languages. This area is poorly studied since in the literature there are almost no works in this direction. We have tried to describe various approaches and achievements of recent years in the development of handwritten recognition models in relation to Cyrillic graphics. The first model uses deep convolutional neural networks (CNNs) for feature extraction and a fully connected multilayer perceptron neural network (MLP) for word classification. The second model, called SimpleHTR, uses CNN and recurrent neural network (RNN) layers to extract information from images. We also proposed the Bluechet and Puchserver models to compare the results. Due to the lack of available open datasets in Russian and Kazakh languages, we carried out work to collect data that included handwritten names of countries and cities from 42 different Cyrillic words, written more than 500 times in different handwriting. We also used a handwritten database of Kazakh and Russian languages (HKR). This is a new database of Cyrillic words (not only countries and cities) for the Russian and Kazakh languages, created by the authors of this work.

show abstract

Hybrid deep learning based automatic speech recognition model for recognizing non-Indian languages

Gupta,

Kumar,

Kumar

2023

Multimed Tools Appl

View full text Add to dashboard Cite

Attention-based Fully Gated CNN-BGRU for Russian Handwritten Text

Cited by 3 publications

References 29 publications

The implications of handwritten text recognition for accessing the past at scale

The implications of handwritten text recognition for accessing the past at scale

Classification of Handwritten Names of Cities and Handwritten Text Recognition using Various Deep Learning Models

Hybrid deep learning based automatic speech recognition model for recognizing non-Indian languages

Contact Info

Product

Resources

About