Siddharth Kumar
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Research Subjects
Siddharth Kumar has a total of 38 co-authors. Here are the top co-authors by number of shared publications:
Jayadeep Pati
Department of Computer Science and Engineering, Indian Institute Of Information Technology Ranchi, Ranchi, Jharkhand, 834010, India
5 shared publications- Jayadeep Pati
Department of Computer Science and Engineering, Indian Institute of Information Technology Ranchi, Ranchi, India
5 shared publications Arghya Chattopadhyay
Banaras Hindu University, India
2 shared publicationsRabinarayan Panda
Department of CSE, Garden City University, Bengaluru, India
2 shared publications- Sujeet Kumar
Indian Institute of Information Technology Ranchi, Ranchi, 834010, India.
2 shared publications - A.N. Bhagat
Department of Computer Science and Engineering, Indian Institute of Information Technology Ranchi, Ranchi, India
1 shared publication
Latest Publications
- Open Access
- 23 July 2026
- Published by Springer Nature in Scientific Reports
- Vol. 16 (1) , 23078
Quantitative correlation of spectroscopic signatures with ligand–protein interactions in anti-cancer drug Afinitor: an integrated experimental–computational study
- 23 July 2026
- Published by Springer Nature in Scientific Reports
- Vol. 16 (1) , 23078
Open AccessQuantitative correlation of spectroscopic signatures with ligand–protein interactions in anti-cancer drug Afinitor: an integrated experimental–computational study
- 2 July 2026
- Published by Springer Nature
- p. 437-452
Hybrid Deep Reinforcement Learning Architectures for Accurate Aircraft Detection in Satellite Imagery
- 2 July 2026
- Published by Springer Nature
- p. 437-452
Hybrid Deep Reinforcement Learning Architectures for Accurate Aircraft Detection in Satellite Imagery
- Open Access
- 18 May 2026
- Published by Springer Nature in Scientific Reports
- Vol. 16 (1)
IBNN: an integrated BERT-neural network framework for sentiment and emotion based text analysis
Social networks are experiencing an unprecedented surge in demand for text mining applications. Text analysis, particularly on online platforms, has become increasingly prevalent. With the rapid growth of social media, enormous volumes of text data, including user comments, are now readily accessible. Analyzing these comments is highly valuable for various business applications. Sentiment Analysis (SA), a natural language processing technique (NLP), is employed to detect emotions and sentiments expressed in text. This paper proposes the development of the IBNN (Integrated BERT with Neural Network) model, which incorporates various feature extraction techniques such as CountVectorizer, TF-IDF Vectorizer, Word2Vec, GloVe, and Bidirectional Encoder Representations from Transformers (BERT) to predict individuals' emotional states through SA. The primary objectives of this work are twofold. First, it compares the effectiveness of these five feature extraction techniques using machine learning, ensemble, and neural network algorithms such as Decision Tree, Random Forest, XGBoost, Neural Network, and the proposed IBNN-across four emotion datasets. Second, it demonstrates the efficacy of the proposed IBNN model, which integrates BERT with neural networks, in achieving highest results across four datasets.
- 18 May 2026
- Published by Springer Nature in Scientific Reports
- Vol. 16 (1)
Open AccessIBNN: an integrated BERT-neural network framework for sentiment and emotion based text analysis
Social networks are experiencing an unprecedented surge in demand for text mining applications. Text analysis, particularly on online platforms, has become increasingly prevalent. With the rapid growth of social media, enormous volumes of text data, including user comments, are now readily accessible. Analyzing these comments is highly valuable for various business applications. Sentiment Analysis (SA), a natural language processing technique (NLP), is employed to detect emotions and sentiments expressed in text. This paper proposes the development of the IBNN (Integrated BERT with Neural Network) model, which incorporates various feature extraction techniques such as CountVectorizer, TF-IDF Vectorizer, Word2Vec, GloVe, and Bidirectional Encoder Representations from Transformers (BERT) to predict individuals' emotional states through SA. The primary objectives of this work are twofold. First, it compares the effectiveness of these five feature extraction techniques using machine learning, ensemble, and neural network algorithms such as Decision Tree, Random Forest, XGBoost, Neural Network, and the proposed IBNN-across four emotion datasets. Second, it demonstrates the efficacy of the proposed IBNN model, which integrates BERT with neural networks, in achieving highest results across four datasets.
- Open Access
- 11 April 2026
- Published by Springer Nature in Scientific Reports
- Vol. 16 (1)
Enhancing Hindi OCR robustness with CRNN-ResNet50: a data augmentation approach on the devanagari dataset
Optical character recognition is a technology that turns texts and scanned documents into digital formats. The practical applications of OCR systems face a lot of challenges because of heterogeneity in scripts, font styles, and different quality images arising during their usage. The proposed work attempts to improve the robustness of OCR models built for the Hindi language by addressing the deficiencies inherent in the widely used IIIT-HW-Dev dataset. Toward this end, two methods of data augmentation are proposed: 1) synthetic images with half characters and conjuncts for complex words, and 2) various image degradations to simulate the real world. The augmented training set is used to train the two deep architectures, namely Convolutional Recurrent Neural Networks (CRNNs) and ResNet-50. Comparisons and in-depth experiments further reveal the merits of the suggested data augmentation techniques: the top performance obtained by the CRNN model exceeds current approaches. On the test set, CRNN achieved a Character Error Rate (CER) of 2.14% and Word Error Rate (WER) of 7.96%, much beyond the CER of 3.27% and WER of 11.82% achieved by the ResNet-50 model. Qualitatively, there was robustness in the case of complex words and images that were also deteriorated. The augmented dataset and trained models can be downloaded by anyone to promote more research in Hindi OCR. So, developing robust OCR systems for complex scripts starts from variability-rich and representative training data along with specific architectures, such as the CRNN architecture.
- 11 April 2026
- Published by Springer Nature in Scientific Reports
- Vol. 16 (1)
Open AccessEnhancing Hindi OCR robustness with CRNN-ResNet50: a data augmentation approach on the devanagari dataset
Optical character recognition is a technology that turns texts and scanned documents into digital formats. The practical applications of OCR systems face a lot of challenges because of heterogeneity in scripts, font styles, and different quality images arising during their usage. The proposed work attempts to improve the robustness of OCR models built for the Hindi language by addressing the deficiencies inherent in the widely used IIIT-HW-Dev dataset. Toward this end, two methods of data augmentation are proposed: 1) synthetic images with half characters and conjuncts for complex words, and 2) various image degradations to simulate the real world. The augmented training set is used to train the two deep architectures, namely Convolutional Recurrent Neural Networks (CRNNs) and ResNet-50. Comparisons and in-depth experiments further reveal the merits of the suggested data augmentation techniques: the top performance obtained by the CRNN model exceeds current approaches. On the test set, CRNN achieved a Character Error Rate (CER) of 2.14% and Word Error Rate (WER) of 7.96%, much beyond the CER of 3.27% and WER of 11.82% achieved by the ResNet-50 model. Qualitatively, there was robustness in the case of complex words and images that were also deteriorated. The augmented dataset and trained models can be downloaded by anyone to promote more research in Hindi OCR. So, developing robust OCR systems for complex scripts starts from variability-rich and representative training data along with specific architectures, such as the CRNN architecture.
- Open Access
- 23 March 2026
- Published by Springer Nature in Scientific Reports
- Vol. 16 (1)
Numerical modeling of CZTS based heterostructured solar cell for high efficiency PV performance
This paper deals with the numerical performance evaluation of solar cells based on ZnMgO/CZTS with and without a back-surface field (BSF) layer. ZnMgO is used as a buffer layer due to its non-toxicity and strain-balancing properties at i-ZnO/ZnMgO and ZnMgO/CZTS interfaces. A multilayer structure of ZnO: Al/i-ZnO/ZnMgO/CZTS is considered for the initial analysis. Subsequently, a BSF layer of CZTS material with different optical and electronic properties, called CZTS2, is inserted between the back-contact layer and the existing CZTS layer, called CZTS1, to form a new structure as ZnO: Al/i-ZnO/ZnMgO/CZTS1/CZTS2 to enhance the performance. The performance of the structure with and without the BSF layer was evaluated in terms of various layer parameters such as thickness, band gap, carrier concentration, defect densities, and work function. Then, the effects of operating temperature and the combination of shunt and series resistance on the overall performance were investigated. The simulated results were calibrated with existing experimental data from the literature to validate the work. In the optimized structure with the BSF layer, a maximum efficiency of 23.67% is achieved, which is 4% higher than that of without the BSF layer. The generation and recombination rates of the structures with and without the BSF layer were investigated to understand the reason for the improved efficiency. The results of this work are very promising for the development of high-efficiency, low-cost, and non-toxic CZTS solar cells.
- 23 March 2026
- Published by Springer Nature in Scientific Reports
- Vol. 16 (1)
Open AccessNumerical modeling of CZTS based heterostructured solar cell for high efficiency PV performance
This paper deals with the numerical performance evaluation of solar cells based on ZnMgO/CZTS with and without a back-surface field (BSF) layer. ZnMgO is used as a buffer layer due to its non-toxicity and strain-balancing properties at i-ZnO/ZnMgO and ZnMgO/CZTS interfaces. A multilayer structure of ZnO: Al/i-ZnO/ZnMgO/CZTS is considered for the initial analysis. Subsequently, a BSF layer of CZTS material with different optical and electronic properties, called CZTS2, is inserted between the back-contact layer and the existing CZTS layer, called CZTS1, to form a new structure as ZnO: Al/i-ZnO/ZnMgO/CZTS1/CZTS2 to enhance the performance. The performance of the structure with and without the BSF layer was evaluated in terms of various layer parameters such as thickness, band gap, carrier concentration, defect densities, and work function. Then, the effects of operating temperature and the combination of shunt and series resistance on the overall performance were investigated. The simulated results were calibrated with existing experimental data from the literature to validate the work. In the optimized structure with the BSF layer, a maximum efficiency of 23.67% is achieved, which is 4% higher than that of without the BSF layer. The generation and recombination rates of the structures with and without the BSF layer were investigated to understand the reason for the improved efficiency. The results of this work are very promising for the development of high-efficiency, low-cost, and non-toxic CZTS solar cells.
- Open Access
- 26 August 2025
- Published by Springer Nature in Discover Computing
- Vol. 28 (1) , 1-23
Advancing Bangla text-to-speech synthesis using a VITS-based model with a custom dataset and comprehensive evaluation
- 26 August 2025
- Published by Springer Nature in Discover Computing
- Vol. 28 (1) , 1-23
Open AccessAdvancing Bangla text-to-speech synthesis using a VITS-based model with a custom dataset and comprehensive evaluation
- 22 January 2024
- Published by IGI Global Scientific Publishing
- p. 106-114
Classification of Arsenic in Groundwater Samples Using Extreme Learning Machine and Crow Search Algorithm for Smart Cities
This research aims to classify the arsenic contamination in the groundwater along the banks of river Ganga of Varanasi, India. The groundwater is vital for various purposes, including agriculture and drinking. Groundwater contamination with high levels of arsenic pose a significant health risk. To tackle this problem, the authors build a model for classifying arsenic levels in groundwater samples that incorporates the extreme learning machine (ELM) algorithm and crowd search optimisation (CSO) technique. In the hybrid approach, they initialize the ELM components and randomly assign weights while employing CSO to guide the search for optimal solutions. By classifying new groundwater samples as having high or low arsenic concentrations, the developed model can be used to evaluate the new groundwater samples. The proposed hybrid approach offers a promising solution for monitoring and managing groundwater quality, ensuring a healthier environment for the city's population. Request access from your librarian to read this chapter's full text.
- 22 January 2024
- Published by IGI Global Scientific Publishing
- p. 106-114
Classification of Arsenic in Groundwater Samples Using Extreme Learning Machine and Crow Search Algorithm for Smart Cities
This research aims to classify the arsenic contamination in the groundwater along the banks of river Ganga of Varanasi, India. The groundwater is vital for various purposes, including agriculture and drinking. Groundwater contamination with high levels of arsenic pose a significant health risk. To tackle this problem, the authors build a model for classifying arsenic levels in groundwater samples that incorporates the extreme learning machine (ELM) algorithm and crowd search optimisation (CSO) technique. In the hybrid approach, they initialize the ELM components and randomly assign weights while employing CSO to guide the search for optimal solutions. By classifying new groundwater samples as having high or low arsenic concentrations, the developed model can be used to evaluate the new groundwater samples. The proposed hybrid approach offers a promising solution for monitoring and managing groundwater quality, ensuring a healthier environment for the city's population. Request access from your librarian to read this chapter's full text.
- 13 October 2023
- Published by Springer Nature in Environmental Monitoring and Assessment
- Vol. 195 (11) , 1-20
Exploring temperature dynamics in Madhya Pradesh: a spatial-temporal analysis
Understanding the dynamics of temperature trends is vital for assessing the impacts of climate change on a regional scale. In this context, the present study focuses on Madhya Pradesh state in Central Indian region to explore the spatial-temporal distribution patterns of temperature changes from 1951 to 2021. Gridded temperature data obtained from the Indian Meteorological Department (IMD) in 1° × 1° across the state are utilised to analyse long-term trends and variations in temperature. The Mann-Kendall (MK) test and Sen’s slope (SS) estimator were used to detect the trends, and Pettitt’s test was utilised for change point detection. The analysis reveals significant warming trends in Madhya Pradesh during the study period during specific time frames. The temperature variables, such as the annual mean temperature (Tmean), maximum temperature (Tmax), and minimum temperature (Tmin), consistently increase, with the most pronounced warming observed during winter. The trend analysis reveals that the rate of warming has increased in the past few years, particularly since the 1990s. However, Pettitt’s test points out significant changes in the temperature, with Tmean rising from 25.46 °C in 1951–2004 to 25.78 °C in 2005–2021 (+0.33 °C), Tmax shifting from 45.77 °C in 1951–2010 to 46.24 °C in 2011–2021 (+0.47°C), and Tmin increasing from 2.65 °C in 1951–1999 to 3.19 °C in 2000–2021 (+0.46 °C). These results, along with spatial-temporal distribution maps, shed important light on the alterations and variations in monthly Tmean, Tmax, and Tmin across the area, underlining the dynamic character of climate change and highlighting the demand for methods for adaptation and mitigation.
- 13 October 2023
- Published by Springer Nature in Environmental Monitoring and Assessment
- Vol. 195 (11) , 1-20
Exploring temperature dynamics in Madhya Pradesh: a spatial-temporal analysis
Understanding the dynamics of temperature trends is vital for assessing the impacts of climate change on a regional scale. In this context, the present study focuses on Madhya Pradesh state in Central Indian region to explore the spatial-temporal distribution patterns of temperature changes from 1951 to 2021. Gridded temperature data obtained from the Indian Meteorological Department (IMD) in 1° × 1° across the state are utilised to analyse long-term trends and variations in temperature. The Mann-Kendall (MK) test and Sen’s slope (SS) estimator were used to detect the trends, and Pettitt’s test was utilised for change point detection. The analysis reveals significant warming trends in Madhya Pradesh during the study period during specific time frames. The temperature variables, such as the annual mean temperature (Tmean), maximum temperature (Tmax), and minimum temperature (Tmin), consistently increase, with the most pronounced warming observed during winter. The trend analysis reveals that the rate of warming has increased in the past few years, particularly since the 1990s. However, Pettitt’s test points out significant changes in the temperature, with Tmean rising from 25.46 °C in 1951–2004 to 25.78 °C in 2005–2021 (+0.33 °C), Tmax shifting from 45.77 °C in 1951–2010 to 46.24 °C in 2011–2021 (+0.47°C), and Tmin increasing from 2.65 °C in 1951–1999 to 3.19 °C in 2000–2021 (+0.46 °C). These results, along with spatial-temporal distribution maps, shed important light on the alterations and variations in monthly Tmean, Tmax, and Tmin across the area, underlining the dynamic character of climate change and highlighting the demand for methods for adaptation and mitigation.
- Open Access
- 24 July 2023
- Published by IWA Publishing in Water Science & Technology
- Vol. 88 (3) , 595-614
The machine learning and geostatistical approach for assessment of arsenic contamination levels using physicochemical properties of water
Arsenic contamination in groundwater due to natural or anthropogenic sources is responsible for carcinogenic and non-carcinogenic risks to humans and the ecosystem. The physicochemical properties of groundwater in the study area were determined in the laboratory using the samples collected across the Varanasi region of Uttar Pradesh, India. This paper analyses the physicochemical properties of water using machine learning, descriptive statistics, geostatistical and spatial analysis. Pearson correlation was used for feature selection and highly correlated features were selected for model creation. Hydrochemical facies of the study area were analyzed and the hyperparameters of machine learning models, i.e., multilayer perceptron, random forest (RF), naïve Bayes, and decision tree were optimized before training and testing the groundwater samples as high (1) or low (0) arsenic contamination levels based on the WHO 10 μg/L guideline value. The overall performance of the models was compared based on accuracy, sensitivity, and specificity value. Among all models, the RF algorithm outclasses other classifiers, as it has a high accuracy of 92.30%, a sensitivity of 100%, and a specificity of 75%. The accuracy result was compared to prior research, and the machine learning model may be used to continually monitor the amount of arsenic pollution in groundwater.
- 24 July 2023
- Published by IWA Publishing in Water Science & Technology
- Vol. 88 (3) , 595-614
Open AccessThe machine learning and geostatistical approach for assessment of arsenic contamination levels using physicochemical properties of water
Arsenic contamination in groundwater due to natural or anthropogenic sources is responsible for carcinogenic and non-carcinogenic risks to humans and the ecosystem. The physicochemical properties of groundwater in the study area were determined in the laboratory using the samples collected across the Varanasi region of Uttar Pradesh, India. This paper analyses the physicochemical properties of water using machine learning, descriptive statistics, geostatistical and spatial analysis. Pearson correlation was used for feature selection and highly correlated features were selected for model creation. Hydrochemical facies of the study area were analyzed and the hyperparameters of machine learning models, i.e., multilayer perceptron, random forest (RF), naïve Bayes, and decision tree were optimized before training and testing the groundwater samples as high (1) or low (0) arsenic contamination levels based on the WHO 10 μg/L guideline value. The overall performance of the models was compared based on accuracy, sensitivity, and specificity value. Among all models, the RF algorithm outclasses other classifiers, as it has a high accuracy of 92.30%, a sensitivity of 100%, and a specificity of 75%. The accuracy result was compared to prior research, and the machine learning model may be used to continually monitor the amount of arsenic pollution in groundwater.