Devanagari

About: Devanagari is a research topic. Over the lifetime, 655 publications have been published within this topic receiving 7428 citations. The topic is also known as: Deva nagari & Hindi Script.

Devanagari Handwritten Character Recognition using Convolutional Neural Networks

Abstract: Devanagari is an Indic script and forms a basis for over 100 languages spoken in India and Nepal including Hindi, Marathi, Sanskrit, and Maithili. It consists of 47 primary alphabets, 14 vowels, 33 consonants, and 10 digits. In addition, the letters of the alphabet are modified when a vowel is added to a consonant. There is no capitalization of letters, like Latin languages. The devanagari script consists of consonants and modifiers. This paper presents a system that works on a set of 29 consonants and one modifier. It uses a self-made Devanagari script dataset which comprises of 29 consonants with no header line (Shirorekha) over them. The dataset has 34604 handwritten images. Deep learning techniques are applied to extract features and recognize the characters in an image. Deep Convolutional Neural Network (DCNN) have been incorporated to extract features and classify the input images. Consecutive convolutional layers are used in this process which brings added advantage in the process of extracting higher-level features. The trained model demonstrated an accuracy of 99.65%.

Role of Context in Devanagari Script Recognition

Abstract: Context plays a dominant role in final interpretation of Devanagari text. This is especially true for hand written texts. The upper and lower modifier symbols are rarely recognised correctly due to their incorrct placement, intermingling and/or incorrect segmentation. A Devanagari text recogniser has to associate these modifier symbols with the appropriate characters from the core strip to arrive at a “syntactically” meaningful word by making appropriate substitutions wherever required. This paper presents methodology and results for Devanagari text recognition using script composition contextual information for post-processing the recognised symbols.

Devanagari character recognition towards natural human-computer interaction

Abstract: Human-computer interaction is a growing research area. There are several ways of interaction with the computer. Handwriting has continued to persist as a means of communication and recording information in the day to day life even with the introduction of new technologies. Due to the growth of technology in India, it becomes important to devise ways that allow people to communicate with computer in Indian languages. Hindi being the national language of India, we present a way to communicate with the computer in Hindi or more precisely, 'Devanagari script'. Due to absence of a global font to represent Devanagari characters, it is important that the computer recognizes the characters written by the user in order to interact with him. The algorithm implemented for character recognition first segments the image containing Devanagari text fed to the software into lines, lines to words and words to characters. The obtained characters are then brought down to a standard size. The Kohonen Neural Network based recognizer then comes into action and recognizes the text character by character and provides the output in Unicode format. The network has been designed with no hidden layer to support quick recognition. Apart from text recognition from an image, we also provided the option to recognize individual handwritten characters drawn using a mouse. Such a system provides keyboard less computer interaction. The technique is implemented using Java. The overall recognition rate for a fixed font machine printed characters is 90.26% and for hand written characters, it is 83.33%.

Experimental Evaluation of Deep Learning Models for Marathi Text Classification

Abstract: The Marathi language is one of the prominent languages used in India. It is predominantly spoken by the people of Maharashtra. Over the past decade, the usage of language on online platforms has tremendously increased. However, research on Natural Language Processing (NLP) approaches for Marathi text has not received much attention. Marathi is a morphologically rich language and uses a variant of the Devanagari script in the written form. This works aims to provide a comprehensive overview of available resources and models for Marathi text classification. We evaluate CNN, LSTM, ULMFiT, and BERT based models on two publicly available Marathi text classification datasets and present a comparative analysis. The pre-trained Marathi fast text word embeddings by Facebook and IndicNLP are used in conjunction with word-based models. We show that basic single layer models based on CNN and LSTM coupled with FastText embeddings perform on par with the BERT based models on the available datasets. We hope our paper aids focused research and experiments in the area of Marathi NLP.

Reading in Devanagari: Insights from functional neuroimaging

Abstract: Objectives: The current study used functional MRI (fMRI) to obtain a comprehensive understanding of the neural network underlying visual word recognition in Hindi/Devanagari, an alphasyllabic - partly alphabetic and partly syllabic Indian writing system on which little research has hitherto been carried out. Materials and Methods: Sixteen (5F, 11M) neurologically healthy, native Hindi/Devanagari readers aged 21 to 50 named aloud 240 Devanagari words which were either visually linear - had no diacritics or consonant ligatures above or below central plane of text, e.g. फल, वाहन, or nonlinear - had at least one diacritic and/or ligature, e.g. फल, किरण, and which further included 120 words each of high and low frequency. Words were presented in alternating high and low frequency blocks of 10 words each at 2s/word in a block design, with linear and nonlinear words in separate runs. Word reading accuracy was manually coded, while fMRI images were acquired on a 3T scanner with an 8-channel head-coil, using a T2*-weighted EPI sequence (TR/TE = 2s/35ms). Results: After ensuring high word naming accuracy (M = 97.6%, SD = 2.3), fMRI data analyses (at FDR P < 0.005) revealed that reading Devanagari words elicited robust activations in bilateral occipito-temporal, inferior frontal and precentral regions as well as both cerebellar hemispheres. Other common areas of activation included left inferior parietal and right superior temporal cortices. Primary differences seen between nonlinear and linear word reading networks were in the right temporal areas and cerebellum. Conclusion: Distinct from alphabetic scripts, which are linear in their spatial organization, and recruit a primarily left-lateralized network for word reading, our results revealed a bilateral reading network for Devanagari. We attribute the additional activations in Devanagari to increased visual processing demands arising from the complex visuospatial arrangement of symbols in this ancient script.