Search

Amharic stands as the second most widely spoken Semitic language globally, trailing only Arabic. As a result, it is a highly resourceful language for hardcopy-document collection, and automatic recognition technology is required. Recognizing handwritten Amharic words is a difficult task because of factors such as variations in individual handwriting for the same words, connectivity of two words without spaces, similarities in the shapes of alphabet characters, and noise from scanned images. The Amharic alphabet has a vast character set that adapted most of its characters from the most popular script called Ethiopic script, formerly known as Ge’ez script, and added unique characters. Despite significant advancements in optical character recognition (OCR) research, Amharic script recognition has received less attention. This study employs a deep learning (DL) approach with additive attention to recurrent neural networks (RNN) to achieve precise recognition of handwritten Amharic words. Seven convolutional neural networks (CNN) and two RNN, using a connectionist temporal classification (CTC) strategy, make up this recognition model architecture, enabling efficient recognition through sequential feature extraction. The study addressed deep learning data insufficiency by using augmentation techniques to increase datasets. The study used an original dataset of 12,047 handwritten Amharic words and an augmented dataset of 22,00 images. The developed model achieved an average character error rate (CER) of 2.84% and an average word error rate (WER) of 9.75% for the testing dataset. These results are indeed promising, offering a glimpse into the potential of this attention-based approach for handwritten Amharic word recognition. This research represents a significant step toward bridging the gap in OCR technology for Amharic script and showcases the transformative capabilities of DL in pattern recognition.

Recognizing handwritten text is a challenging task, especially for scripts with numerous alphabets and symbols. The Ethiopic script has a vast character set and is used for historical documents in typewritten, handwritten, and hand-printed forms. However, despite its importance as an ancient script, optical character recognition research has not given enough attention to Ethiopic text recognition. In recent years, deep learning (DL) has emerged as a powerful technique for recognizing patterns. In this study, a DL approach is used to recognize historical Ethiopic handwritten texts. The recognition model employs an end-to-end strategy enabling sequential feature extraction and efficient recognition. An attention mechanism coupled with a connectionist temporal classification architecture is the core of this recognition model architecture. In addition, there are seven convolutional neural networks and two recurrent neural networks. We increase the training data using data augmentation techniques to address the data scarcity common in deep learning applications. The experiments include an original training dataset of 79,684 historical handwritten images and an augmented dataset of 10,000 images containing Ethiopic texts. The model used for recognition showed promising results. For “Test Set I” which had 6,150 samples, the character error rate (CER) was 17.95%, and for “Test Set II” which had 15,935 samples, the CER was 29.95%. These outcomes indicate that this approach has the potential to improve the recognition of historical handwritten Ethiopic text.