Volume 13 | Issue 4 |

 Abstract

This project proposes an AI-based Petition Analysis and Grievance Management System that automates public complaint handling. The system, employing NLP and ML, categorizes petitions, identifies urgency, and directs them to the right departments. Dashboards for real-time tracking promote transparency and accountability, while sentiment analysis prioritizes crucial issues. Manual effort is minimized, response time is enhanced, and data-driven governance is enabled through actionable insights into public concerns.

Licence: creative commons attribution 4.0

  License

Creative Commons Attribution 4.0 and The Open Definition

 Keywords

AI, Machine Learning, Natural Language Processing, Grievance Redressal,

  License

Creative Commons Attribution 4.0 and The Open Definition

 Abstract

This dissertation examines the changing nexus of Artificial Intelligence (AI) and corporate legal practice in India, providing a detailed analysis of how AI technologies are reconfiguring the functioning, delivery, and regulation of legal services in the corporate space. As India undergoes a rapid digitalization across industries, the legal sector--historically considered conservative and process-oriented--is increasingly adopting AI-led innovations to boost efficiency, precision, and decision-making. The research commences by situating the worldwide rise of AI technologies and chronicles their development in legal frameworks, with specific reference to the Indian business legal context. It discusses the implementation of AI-based tools across the most significant legal procedures like contract analysis, legal research, due diligence, litigation planning, fraud detection, and regulatory compliance. By citing particular platforms such as CaseMine, Prarambh (formed by Cyril Amarchand Mangaldas), Anuvaad, and global systems such as IBM Watson and COIN by JPMorgan Chase, the research brings forth the real-world application of AI within corporate law practice. Using doctrinal and comparative legal research approaches, the dissertation examines the role of AI in improving speed, lowering costs, and enhancing risk mitigation in corporate legal processes. It also considers the law and ethics aspects of AI embedding--data protection issues, prejudice through algorithms, professional negligence, and erosion of human judicial wisdom. The legislative framework is viewed critically in light of a comparison of legal instruments and governmental interventions across the United Kingdom, United States, European Union, China, and Australia, and offers India's path to regulation valuable lessons. An important value added to this work is the in-depth analysis of Indian legal laws--that include the Companies Act, 2013; SEBI legislation; the Information Technology Act, 2000; and incoming data protection acts--and how these intersect with applications of AI for legal purposes. The research ends on a note proposing a strategic map for the Indian legal profession and suggesting regulation amendments, moral benchmarks, and professionalism guidelines in place to see the use of AI in the corporate legal fraternity used responsibly, fairly, and openly. In conclusion, this dissertation presents a timely and forward-looking analysis of the ways in which AI can enhance, supplement, and possibly change legal practice within India's corporate world, such that technological development is in tune with constitutional principles, client interest, and the fundamental values of justice.

Licence: creative commons attribution 4.0

  License

Creative Commons Attribution 4.0 and The Open Definition

 Keywords

Artificial Intelligence (AI), Corporate Legal Sector, Legal Technology, AI in Indian Law, Legal Research Automation, Contract Analysis, Due Diligence, Compliance Monitoring, Litigation Management, Companies Act 2013, SEBI Regulations, Legal Ethics, Algorithmic Bias, AI Governance,

  License

Creative Commons Attribution 4.0 and The Open Definition

 Abstract

Knee osteoarthritis (KOA) is a degenerative joint condition that affects millions globally, especially older adults. Timely and accurate diagnosis is essential to slow disease progression. This paper presents a deep learning-based system for automated KOA detection using X-ray images, graded according to the Kellgren and Lawrence (KL) scale. Four convolutional neural networks--ResNet-34, VGG-19, DenseNet-121, and DenseNet-161--are fine-tuned through transfer learning and combined using an ensemble strategy. To model the ordered nature of KOA severity, Conditional Ordinal Regression (CORN) is employed. The system integrates Explainable AI (XAI) using Eigen-CAM visualizations to highlight diagnostic regions in the X-ray images. Evaluation on the Osteoarthritis Initiative dataset shows state-of-the-art results, with 98% accuracy and a Quadratic Weighted Kappa (QWK) score of 0.99. The final model is deployed via a Streamlit web application, offering an accessible interface for real-time diagnosis. The approach provides a reliable and interpretable tool for assisting radiologists in KOA assessment.

Licence: creative commons attribution 4.0

  License

Creative Commons Attribution 4.0 and The Open Definition

 Keywords

Knee Osteoarthritis, Deep Learning, Kellgren-Lawrence Grading, Explainable AI.

  License

Creative Commons Attribution 4.0 and The Open Definition

 Abstract

In modern educational institutions, effective communication is a key pillar of administrative efficiency. Traditional notice boards, often paper-based and manually updated, pose significant limitations in terms of scalability, timeliness, and environmental sustainability. This paper presents the design and implementation of an AI-powered smart notice board system that addresses these challenges through automation, multimedia integration, and conversational AI. The proposed system is built around a Raspberry Pi 4 platform, functioning as a compact and affordable local server. It hosts a Django-based web application that allows authorized administrators to upload and manage notices in the form of text, images, and videos. These notices are dynamically rendered on a connected HDMI display in a continuous loop. The system is further enhanced by the integration of a Rasa-powered chatbot, embedded within the display interface, which enables real-time interaction with users. The chatbot is trained to handle frequently asked academic queries, including examination schedules, project deadlines, and placement updates, thereby reducing repetitive student-faculty interactions. Designed to operate fully offline, the system is ideal for deployment in environments with limited network infrastructure. It emphasizes paperless communication, user interactivity, and real-time responsiveness. Extensive testing confirms the system's stability, ease of use, and potential for scalability across departments and institutions. This work contributes to the ongoing digital transformation of educational infrastructure, combining IoT, web technologies, and natural language processing into a cohesive smart campus solution.

Licence: creative commons attribution 4.0

  License

Creative Commons Attribution 4.0 and The Open Definition

 Keywords

Smart Notice Board, Raspberry Pi, Django, Rasa Chatbot, Artificial Intelligence.

  License

Creative Commons Attribution 4.0 and The Open Definition

 Abstract

Adverse Drug Reactions (ADRs) resulting from drug-drug interactions are a major healthcare concern. While Graph Neural Networks (GNNs) effectively model these interactions, their one-dimensional processing limits complex feature extraction. This research introduces a novel extension by integrating a two-dimensional Convolutional Neural Network (CNN2D) to enhance ADR prediction. By converting drug interaction data into 2D matrices, CNN2D captures intricate spatial relationships, complementing the GNN's graph-based insights. This hybrid model achieves a superior prediction accuracy of 99.87%, significantly outperforming traditional methods like KNN and Decision Trees. The extension showcases the power of deep learning in advancing drug safety evaluation.

Licence: creative commons attribution 4.0

  License

Creative Commons Attribution 4.0 and The Open Definition

 Keywords

Adverse Drug Reactions, Drug-Drug Interactions, Graph Neural Networks, Convolutional Neural Networks, Self-Supervised Learning, SMILES Representation, Deep Learning, Side Effect Prediction, Drug Safety, TF-IDF Vectorization.

  License

Creative Commons Attribution 4.0 and The Open Definition

 Abstract

The growing burden of Non-Performing Assets (NPAs) in Indian Public Sector Banks (PSBs) poses a significant threat to financial stability and economic growth. This article evaluates the effectiveness of current legal and institutional mechanisms for NPA resolution, such as SARFAESI, DRTs, and the Insolvency and Bankruptcy Code (IBC). Despite their roles, persistent issues like judicial delays, enforcement gaps, and inadequate institutional coordination hinder their success. Through critical evaluation and comparison with global practices, this article proposes a strategic shift from reactive recovery to proactive reforms aimed at sustainable resolution.

Licence: creative commons attribution 4.0

  License

Creative Commons Attribution 4.0 and The Open Definition

 Keywords

NPAs, Public Sector Banks, SARFAESI, IBC, DRT, Legal Reform, Sustainable Finance, Banking Law

  License

Creative Commons Attribution 4.0 and The Open Definition

 Abstract

Background: Suicide among adolescents is a rising global concern, particularly in low- and middle-income countries. Adolescents often face stressors that predispose them to suicidal ideation and behaviour. Objective: To evaluate the effectiveness of a structured teaching program in improving knowledge about risk factors and prevention of suicidal behaviour among adolescents. Methods: A pre-experimental one-group pre-test post-test design was used. A total of 100 adolescents aged 10-16 years from selected schools in Bangalore Urban were selected through non-probability convenience sampling. A structured self-administered questionnaire was used to assess knowledge before and after the intervention. Results: In the pre-test, 67% had inadequate knowledge, 33% had moderate knowledge, and none had adequate knowledge. In the post-test, 76% had adequate knowledge, 24% had moderate knowledge, and none remained in the inadequate category. A significant increase in mean knowledge score was observed (pre-test: 8.35, post-test: 16.31), with a mean difference of 7.96 (t=31.09, p<0.05). Conclusion: The structured teaching program was effective in enhancing adolescents' knowledge regarding suicidal risk factors and preventive measures. Keywords: adolescent mental health, suicide prevention, structured teaching program, risk factors, nursing education.

Licence: creative commons attribution 4.0

  License

Creative Commons Attribution 4.0 and The Open Definition

 Keywords

Keywords: adolescent mental health, suicide prevention, structured teaching program, risk factors, nursing education.

  License

Creative Commons Attribution 4.0 and The Open Definition

 Abstract

Effective detection has become a critical challenge due to the rise of IoT devices, which has led to an increase in botnet attacks. This research extends traditional botnet detection models by incorporating advanced ensemble deep learning techniques to improve prediction accuracy. We integrate CNN, LSTM, and GRU in hybrid architectures such as CNN + LSTM + GRU and CNN + BiLSTM + GRU, which effectively capture both spatial and temporal patterns in IoT network traffic. Feature selection using Mutual Information optimises model performance, reducing computational complexity while improving detection efficiency. Additionally, a Flask is used to create a user-friendly front-end application, which allows for smooth testing and evaluation of the model. Secure user authentication protects sensitive information and ensures data integrity. The experiment's findings demonstrate that the suggested ensemble models achieve superior accuracy, surpassing 97%, in detecting botnet activity, highlighting their effectiveness in securing IoT environments.

Licence: creative commons attribution 4.0

  License

Creative Commons Attribution 4.0 and The Open Definition

 Keywords

Botnet Detection, IoT Security, Deep Learning, CNN, LSTM, GRU, Hybrid Models, Ensemble Learning, Feature Selection, Mutual Information, Flask Framework, User Authentication, Cybersecurity.

  License

Creative Commons Attribution 4.0 and The Open Definition

 Abstract

A common dental ailment called periodontitis is brought on by bacterial infection of the tooth's surrounding bone. To avoid serious consequences like tooth loss, early identification and accurate treatment are essential. Dental experts have historically diagnosed periodontal disease by manually identifying and labelling the condition, a procedure that takes a great deal of skill and involves tedious, time-consuming activities. The goal of this work is to use dental imaging datasets to automatically detect and classify periodontitis by utilising sophisticated neural network architectures. By effectively analysing photos for early-stage illness detection using deep learning techniques, the suggested method lessens the need for manual inspection. Multiple optimisation tactics inside the neural networks are compared to show how they affect detection performance. Results reveal that the suggested technique provides greater accuracy, with a 2D Convolutional Neural Network model having a detection accuracy of 96.93%. This high-performance solution highlights the promise of automated systems in strengthening diagnostic precision, efficiency, and scalability for periodontitis, thereby improving patient outcomes and streamlining clinical procedures.

Licence: creative commons attribution 4.0

  License

Creative Commons Attribution 4.0 and The Open Definition

 Keywords

YOLOv8; Tooth Position Detection; Periodontitis; Bitewing Radiograph; Convolutional Neural Networks (CNN); Deep Learning; Dental Imaging; Automated Diagnosis; Medical Image Processing; Early Disease Detection; Diagnostic Accuracy; Neural Network Optimization.

  License

Creative Commons Attribution 4.0 and The Open Definition

 Abstract

In today's era of information overload, accessing relevant and meaningful news has become increasingly challenging. This research presents a Personalized News Aggregator with Sentiment Analysis--a system designed to deliver user-centric news content tailored to individual interests and preferences. The platform aggregates news from diverse sources and leverages Natural Language Processing (NLP) techniques to analyze the sentiment of each article, helping users better understand the emotional tone and context of the information they consume. The system integrates machine learning models for sentiment analysis with recommendation algorithms to enable personalized content delivery. By offering features such as filtering, keyword search, and sentiment-based categorization, the solution addresses limitations found in traditional news platforms. This paper explores the technical implementation of the system, including data collection, preprocessing, model selection, and web-based deployment. It also highlights the system's potential in enhancing information accessibility and improving user satisfaction through a more refined, relevant, and engaging news experience.

Licence: creative commons attribution 4.0

  License

Creative Commons Attribution 4.0 and The Open Definition

 Keywords

Personalized News Aggregator, Sentiment Analysis, Natural Language Processing (NLP), Machine Learning, Recommendation Algorithms, User Preferences, News Personalization.

  License

Creative Commons Attribution 4.0 and The Open Definition