SBSantosh Kumar Bharti

Santosh Kumar Bharti

*This is an algorithmically generated author record based on publicly available information.

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


Santosh Kumar Bharti has a total of 69 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    CertAV: Manifold-Aware Certified Robustness for Audio-Visual Deepfake Detection

    by Palak Parmar, Chintan Bhatt, SantoshKumar Bharti, Tanvish Desai

  • Open Access

    Extractive Text Summarization Using Formality of Language

    by Harsh Mehta, Santosh Kumar Bharti, Nishant Doshi

    Automatic text summarization has been a prominent research topic for over a decade, aiming to distill concise summaries from extensive textual documents. This study introduces a novel approach addressing the intricacies of morphologically rich Indo-Iranian languages. We propose a unique method that leverages linguistic formality to guide summary generation. Building on an existing formality formula designed for English, we adapt and extend it for the structural characteristics of Indo-Iranian languages, which follow the Subject-Object-Verb (SOV) order. Our refined formula demonstrates a 7.28% improvement in formality scores compared to informal texts, validated through statistical significance testing. To assess sentence formality, we use our custom formula alongside additional features such as Shannon entropy scores and numeric token presence, combining these into a comprehensive sentence evaluation metric. Using this framework, we generate extractive summaries of Gujarati texts. Comparative evaluations at 20% and 30% compression ratios reveal that our method outperforms existing baselines, with ROUGE-1 score improvements of 14.63% at 30% and 28.60% at 20% compression. For reproducibility and further exploration, all experimental data and source code are made publicly available.


  • Open Access

    Extractive Text Summarization Using Formality of Language

    by Harsh Mehta, Santosh Kumar Bharti, Nishant Doshi

    Automatic text summarization is a more than decade-old research method to get a summary from the larger descriptive text, and the cam- paign to find the best approach is still going on. Much research has been done, and new techniques are emerging to get a summary from the text document. Many unsupervised methods do not give justice to the lan- guage with high inflections. We aim to provide a unique technique that gives justice to complex morphological language. We present a novel method, the formality of a language, to generate a summary. We have used an already researched formality formula to measure the formality of language but customized it to suit an Indo-Iranian language family. We have proved that our redefined formula works with a 7.28% higher formality score than informal language. We also ran a statistical sig- nificance test to confirm the results. We apply this formula to measure whether the sentence is formal or informal. Our customized formula mix- ture with other measures like Shannon entropy score, and the presence of numeric values, we generate a score of a sentence. Based on scoring, we generate a text summary. We experimented formula on the generated Gujarati corpus and collected the results. We compared our technique with similar methods on ideal compression ratios like 20% and 30%. We got the highest accuracy with an increase of 14.63% compared to other methods on rouge-1 with a 30% compression ratio and a 28.60% increase in rouge-1 with a 20% compression ratio. Detailed experiment values with all the results have been provided for further study and research.

  • Text Summarization Corpus Generation in Low-Resource South Asian Gujarati Language

    by Harsh Mehta, Santosh Kumar Bharti, Nishant Doshi

    The scarcity of Datasets/Corpus belonging to resource-poor languages is a widely discussed topic. Due to the absence of widespread usage of the local language in Online platforms, the English language is dominant. Much research comes to a standstill without a proper dataset. Due to the need for more appropriate datasets in the Gujarati language, we have made a Gold standard dataset for multiple uses, mainly for Text summarization, with 76% Cohen's Kappa Inter-Annotator agreement, which is a satisfactory level to approve the dataset for the presented hypothesis. Also, the TF - IDF score for the baseline system can be helpful in further research. We have also measured the formality score of the annotator's text. For annotator-1, it received an 82.38 % formality score, and for annotator-2, it received an 81.79 % formality score, meaning annotators selected more formal statements and conveyed more information. We also checked the formality score result with the Wilcoxon Signed-Rank Test with a p-value of 0.438, which is quite acceptable.


  • Deep Learning-Based Framework for URL Phishing Detection

    by Meet Kava, Rajeev Kumar Gupta, Santosh Bharti, Abhijit Kumar


  • Deep Learning-Based Agriculture Monitoring and Forecasting

    by Ishaan Mehta, Santosh Kumar Bharti, Honey Mehta, Rajeev Kumar Gupta

    Smart agriculture has become crucial in meeting the increasing dietary needs of a growing population, particularly in countries where agriculture has significant economic impact. Deep learning techniques, such as convolutional neural networks and recurrent neural networks, have been extensively researched and applied in agriculture in recent years. In this study, recent research articles on deep learning in agriculture over the past five years are analyzed to identify key contributions and challenges. The study has also explored agriculture parameters monitored by the internet of things and used them to train the deep learning algorithms for analysis. The study compares various factors across different studies, including the agriculture area of focus, dataset used, deep learning model and framework, data preprocessing and augmentation methods, and accuracy of results.


  • AI-Based Plant Disease Detection and Classification Using Pretrained Models

    by Honey Mehta, Rajeev Kumar Gupta, Arti Jain, Santosh Kumar Bharti, Nilesh Kunhare

    Plants growth is crucial to an agricultural industry, and to the economy of a nation. Consequently, taking care of plants is essential. Like humans, plants are susceptible to several bacterial, fungal, and viral diseases. To avoid the plants from being destroyed, prompt disease detection and treatment are crucial. The goal of this chapter is to introduce plant disease detection and classification using deep learning and transfer learning based pre-trained models. Using images of leaves, the models can identify several plant ailments. The models, namely- Convolutional Neural Network (CNN), MobileNet, and VGG16 are applied for plant disease identification. To train them, the PlantVillage dataset is used, and to enhance the sample size, the dataset is augmented. For experimentation purposes, images of both healthy and damaged plants are taken. Experiment results reveal that VGG16 has outperformed CNN and MobileNet models for the detection of tomato, potato, and apple plant diseases. The accuracy of the VGG16 model is 0.89, 0.92, and 0.95 for the tomato, potato, and apple plant diseases respectively.