Handwritten Word Recognition using Deep Learning Approach: A Novel Way of Generating Handwritten Words

Akter, Mst Shapna; Shahriar, Hossain; Cuzzocrea, Alfredo; Ahmed, Nova; Leung, Carson K.

doi:10.1109/bigdata55660.2022.10021025

Cited by 8 publications

(5 citation statements)

References 45 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Also, the authors in [27] a method for generating diverse handwritten word images using handwritten characters is proposed. The approach involves training a BiLSTM-CTC architecture with synthetic handwritten words generated in two ways: overlapped and non-overlapped.…”

Section: Results Discussionmentioning

confidence: 99%

Real time handwriting recognition system using CNN algorithms

Maryam Yaseen Abdullah

2023

WJCMS

View full text Add to dashboard Cite

Abstract— The growing use of digital technologies across various sectors and daily activities has made handwriting recognition a popular research topic. Despite the continued relevance of handwriting, people still require the conversion of handwritten copies into digital versions that can be stored and shared digitally. Handwriting recognition involves the computer's strength to identify and understand legible handwriting input data from various sources, including document, photo-graphs and others. Handwriting recognition pose a complexity challenge due to the diversity in handwriting styles among different individuals especially in real time applications. In this paper, an automatic system was designed to handwriting recognition using the recent artificial intelligent algorithms, the conventional neural network (CNN). Different CNN models were tested and modified to produce a system has two important features high performance accuracy and less testing time. These features are the most important factors for real time applications. The experimental results were conducted on a dataset includes over 400,000 handwritten names; the best performance accuracy results were 99.8% for SqueezeNet model.

show abstract

Section: Results Discussionmentioning

confidence: 99%

Real time handwriting recognition system using CNN algorithms

Maryam Yaseen Abdullah

2023

WJCMS

View full text Add to dashboard Cite

show abstract

“…A graphical comparison with earlier research is shown in Fig. 10, and a comparison of the suggested method's performance metrics with those of Alonso et al [3], Gan et al [4], Luo et al [5], Fogel et al [6], Akter et al [9] and Kang et al [14] is shown in Table 4.…”

Section: Evaluation Of the Approach Against Prior Workmentioning

confidence: 87%

“…Wang et al [8] highlighted the benefits of transfer learning in speech and language processing. In order to improve handwritten word recognition using synthetic datasets, Akter et al provided a deep-text-recognition-benchmark (BiLSTM-CTC) based technique that shows promise in tackling data scarcity, especially for languages like Bangla [9]. A fresh assessment metric was introduced by Tüselmann et al [10] to solve problems in semantic word recognition.…”

Section: Literature Surveymentioning

confidence: 99%

A Harmonic Approach to Handwriting Style Synthesis Using Deep Learning

Tusher,

Kongara,

Pande

et al. 2024

CMC

View full text Add to dashboard Cite

The challenging task of handwriting style synthesis requires capturing the individuality and diversity of human handwriting. The majority of currently available methods use either a generative adversarial network (GAN) or a recurrent neural network (RNN) to generate new handwriting styles. This is why these techniques frequently fall short of producing diverse and realistic text pictures, particularly for terms that are not commonly used. To resolve that, this research proposes a novel deep learning model that consists of a style encoder and a text generator to synthesize different handwriting styles. This network excels in generating conditional text by extracting style vectors from a series of style images. The model performs admirably on a range of handwriting synthesis tasks, including the production of text that is out-of-vocabulary. It works more effectively than previous approaches by displaying lower values on key Generative Adversarial Network evaluation metrics, such Geometric Score (GS) (3.21 × 10 −5 ) and Fréchet Inception Distance (FID) (8.75), as well as text recognition metrics, like Character Error Rate (CER) and Word Error Rate (WER). A thorough component analysis revealed the steady improvement in image production quality, highlighting the importance of specific handwriting styles. Applicable fields include digital forensics, creative writing, and document security.

show abstract

“…The number of hate samples in each category is imbalanced. Akter et al (2022) [10] have classi ed aggressive comments in Hindi, Bangla, and English datasets using LSTM, BiLSTM, LSTMAutoencoder, word2vec, BERT, and GPT-2 models and have also shown a novel way of generating machine-translated data to resolve data unavailability issues. The BERT model performed best on noisy datasets, with 78% accuracy, while the GPT2 model performed best on raw datasets that did not contain any noise, with 80% accuracy.…”

Section: Related Workmentioning

confidence: 99%

Detection of Banglish Slang in Social Media Comments Using a Hybrid Bidirectional Long Short-Term Memory (Bi-LSTM) Model

Khalaf,

Yasmin,

Jisan

et al. 2024

Preprint

View full text Add to dashboard Cite

Banglish slang is a distinctive mashup of Bengali and English terms that is widely used in messaging applications, social media, and other casual contexts. It has grown to be an essential component of Bangladesh's linguistic landscape. The use of Banglish slang in social media comments has become more common among young people from Bangladesh in recent years. Yet, because traditional language processing methods are not appropriate for managing such linguistic complexity, automated text analysis systems find it difficult to handle this informal communication style. Our proposed model captures the unique components of Banglish slang, such as slang phrases, misspellings, and abbreviations, by combining character-level and word-level properties since it is the most common form of communication and improves user interaction and content moderation. We remove special characters from the dataset and transform all test results to lowercase. After then, the stop words are removed from the dataset. Furthermore, we do tokenization using a pad sequence prior to classification. Stack Bidirectional Long Short-Term Memory (Bi-LSTM) model, a neural network architecture used in Natural Language Processing (NLP), is a state-of-the-art model utilized in this research to classify slang in Banglish words. On the training set with an outstanding 99% accuracy and a strong 79.8% validation accuracy, the suggested model beats traditional approaches. The Mendeley dataset was used to analyze social media comments to detect instances of Banglish slang. To carry out the work, approximately five thousand comments are considered overall. The experimental findings show how well our recommended approach works to identify Banglish slang. This method can be useful in monitoring content and language learning, among other real-world applications.

show abstract

Handwritten Word Recognition using Deep Learning Approach: A Novel Way of Generating Handwritten Words

Cited by 8 publications

References 45 publications

Real time handwriting recognition system using CNN algorithms

Real time handwriting recognition system using CNN algorithms

A Harmonic Approach to Handwriting Style Synthesis Using Deep Learning

Detection of Banglish Slang in Social Media Comments Using a Hybrid Bidirectional Long Short-Term Memory (Bi-LSTM) Model

Contact Info

Product

Resources

About