Volume 14 | Issue 4 |

 Abstract

Recommendation systems are integral to contemporary e-commerce platforms, facilitating user discovery of relevant products amidst vast inventories. This paper offers a comprehensive review and analysis of various recommendation methodologies employed within e-commerce environments. It examines diverse approaches, including collaborative filtering, association rule mining, clustering algorithms, ensemble learning, and deep learning models. The study critically evaluates multiple research works, detailing their methodologies, findings, and inherent limitations. Based on this analysis, the paper underscores the significance of hybrid machine learning models that integrate several recommendation techniques to enhance personalization and predictive accuracy. The findings suggest that incorporating behavioral data, sentiment analysis, and advanced deep learning methods can substantially improve the efficacy of recommendation systems in large-scale e-commerce platforms.

Licence: creative commons attribution 4.0

  License

Creative Commons Attribution 4.0 and The Open Definition

 Keywords

Recommendation System, E-Commerce Platforms, Machine Learning, Collaborative Filtering, Deep Learning, Personalization

  License

Creative Commons Attribution 4.0 and The Open Definition

 Abstract

Large Language Models (LLMs) have advanced data analysis capabilities, but most systems rely on cloud infrastructure, raising concerns about privacy, latency, and data security. This paper presents an offline-first data analysis system that integrates locally deployed LLMs with a Retrieval-Augmented Generation (RAG) model to enable secure and context-aware insights without cloud dependency. The proposed system processes data entirely on local machines using document ingestion, vector retrieval, and LLM-based reasoning, ensuring that sensitive information remains within the user environment. Additionally, federated privacy principles are incorporated to support decentralized and secure model updates. The system is designed for low-spec devices and supports real-time querying, document understanding, and knowledge synthesis. This paper discusses the architecture, implementation, and performance trade-offs, demonstrating a scalable and privacy-preserving solution for intelligent data analysis in offline environments.

Licence: creative commons attribution 4.0

  License

Creative Commons Attribution 4.0 and The Open Definition

 Keywords

Local LLM, Retrieval-Augmented Generation (RAG), Offline AI, Federated Privacy, Edge Computing, Privacy-Preserving Analytics

  License

Creative Commons Attribution 4.0 and The Open Definition

 Abstract

Skin aging is a complex biological process influenced by intrinsic factors such as genetic predisposition and hormonal changes, as well as extrinsic factors including ultraviolet (UV) radiation, environmental pollution, and lifestyle habits. These factors lead to the formation of reactive oxygen species (ROS), which cause oxidative stress and damage to cellular structures, resulting in visible signs of aging such as wrinkles, fine lines, dryness, and loss of elasticity. The present study aims to formulate and evaluate a polyherbal anti-aging gel containing multiple herbal extracts with antioxidant and skin-protective properties. The formulation includes Zingiber zerumbet, Calendula officinalis, Glycyrrhiza glabra (licorice), Camellia sinensis (green tea), and Aloe vera. Beetroot extract was incorporated as a natural coloring agent, while methyl paraben was used as a preservative. Carbopol 940 was employed as a gelling agent, and triethanolamine was used for neutralization and pH adjustment. The prepared gel was evaluated for various physicochemical parameters including pH, viscosity, spreadability, extrudability, and homogeneity. Additionally, UV-visible spectrophotometric analysis and microbiological evaluation were performed. The results indicated that the formulation exhibited acceptable pH, good viscosity, excellent spreadability, and no microbial growth. The presence of multiple herbal extracts provided synergistic antioxidant activity, making the formulation effective for anti-aging purposes. Thus, the developed polyherbal gel can be considered a safe, stable, and effective alternative to synthetic anti-aging products.

Licence: creative commons attribution 4.0

  License

Creative Commons Attribution 4.0 and The Open Definition

 Keywords

Polyherbal gel, Anti-aging, Antioxidants, Zingiber zerumbet, Herbal cosmetics

  License

Creative Commons Attribution 4.0 and The Open Definition

 Abstract

Licence: creative commons attribution 4.0

  License

Creative Commons Attribution 4.0 and The Open Definition

 Keywords

Truc Lam Zen; Emperor-Monk Tran Nhan Tong; Scriptural Landscape; Post-Ascetic Space; Cultural Memory; Vietnam Buddhism; Yen Tu

  License

Creative Commons Attribution 4.0 and The Open Definition

 Abstract

This study reports taxonomic characterization of seven species of Curvularia, viz. C. kusanoi, C. trifolii, C. coatesiae, C. falsilunata, C. lunata, C. umbiliciformis and C. phaespora, three species of Epicoccum, viz. E. sorghinum, E. latusicollum and E. nigrum, two species of Nigrospora viz. N. sphaerica and N. oryzae among others isolated as endophytic fungi, from grasses of Kodagu district in Karnataka state, India. These endophytes were identified by 18S rDNA sequencing protocol

Licence: creative commons attribution 4.0

  License

Creative Commons Attribution 4.0 and The Open Definition

 Keywords

Grasses, endophytic fungi, molecular identification, 18S rDNA

  License

Creative Commons Attribution 4.0 and The Open Definition

 Abstract

With the rapid proliferation of digital messaging platforms, users face increasing exposure to cyber threats including phishing attacks, fraudulent links, misleading information, and fake product promotions. Traditional security solutions are largely reactive, relying on post-event analysis and basic spam filtering, which proves insufficient against sophisticated real-time threats. This paper presents Parasite AI, an intelligent real-time chat monitoring assistant that leverages Natural Language Processing (NLP) and machine learning techniques to continuously analyze visible chat content and detect potential threats during live conversations. The proposed system employs Android accessibility services to capture on-screen chat data without interrupting user workflows, processes the extracted text through an AI analysis engine incorporating keyword pattern matching, contextual intent recognition, and URL safety verification, and delivers real-time alerts and intelligent suggestions via a non-intrusive overlay interface. The system integrates multiple specialized modules including screen monitoring, text extraction, AI-based threat analysis, link verification, claim detection, and product authentication to provide comprehensive protection. Experimental evaluation across diverse messaging scenarios demonstrates that the system achieves 95.2% threat detection accuracy with a false positive rate of 4.8%, while maintaining sub-second response times. The system's modular architecture supports seamless integration with popular messaging platforms including WhatsApp, Telegram, and social media applications, establishing Parasite AI as a practical and scalable solution for enhancing user safety in digital communication environments

Licence: creative commons attribution 4.0

  License

Creative Commons Attribution 4.0 and The Open Definition

 Keywords

Real-time chat monitoring Natural language processing Phishing detection Overlay interface Accessibility services Cybersecurity AI

  License

Creative Commons Attribution 4.0 and The Open Definition

 Abstract

Deaf and mute communities include individuals with partial or complete hearing and speech impairments who primarily depend on sign language for communication. However, interaction with non-signers remains difficult due to limited sign language awareness, creating barriers in education, healthcare, workplaces, and public services. Existing assistive communication systems face challenges such as restricted vocabulary coverage, reliance on predefined gestures, poor real-time performance, limited support for regional sign languages like Indian Sign Language, lack of two-way communication, and inadequate handling of continuous gestures, facial expressions, and environmental noise, which significantly reduces their real-world effectiveness. To overcome these challenges, this project introduces an intelligent, AI-driven communication framework designed to provide seamless two-way interaction between sign-language users and non-signers. The proposed system incorporates a Transformer-Based Gesture Encoder that analyzes continuous hand movements, facial expressions, and body dynamics from live video streams to accurately recognize Indian Sign Language. This enables natural and real-time interpretation of signs without relying on predefined gesture templates, thereby improving recognition accuracy and adaptability in dynamic environments. In addition, the system integrates a Speech Recognition and Synthesis Module utilizing advanced techniques such as RNN-Transducer, Connectionist Temporal Classification, and Deep Neural Networks to convert spoken language into text with high reliability, even under noisy conditions. The interpreted text is further processed by an Avatar Module based on Neural Avatar Synthesis, which generates realistic, human-like sign language animations to visually convey spoken or typed content. Implemented through a web-based interface, the system ensures accessibility, scalability, and ease of use, ultimately promoting inclusive and effective communication across diverse user groups.

Licence: creative commons attribution 4.0

  License

Creative Commons Attribution 4.0 and The Open Definition

 Keywords

Neural 3D Avatar, Bidirectional Communication, Deaf and Hearing Users, Sign Language Recognition, Sign Language Translation, Gesture Recognition, Computer Vision, Deep Learning, Human-Computer Interaction, Speech-to-Sign Conversion, Sign-to-Speech Conversion, Real-Time Communication, AI-based Avatar, Natural Language Processing, Assistive Technology

  License

Creative Commons Attribution 4.0 and The Open Definition

 Abstract

Job satisfaction and employee productivity are closely connected in any organization. When employees feel happy and valued, they tend to perform better at work. Many factors like salary, work freedom, and recognition influence how satisfied employees feel. Understanding these factors helps organizations improve overall performance. A positive work environment also supports better productivity and employee morale. The main objective of this study is to examine the relationship between job satisfaction and employee productivity in the organization. The study further assesses employee satisfaction with salary and compensation, understands employee views on job autonomy, evaluates recognition and feedback practices, and measures overall employee productivity. A descriptive research design and convenience sampling method have been used with a sample size of 110 respondents. Primary data was collected through a structured questionnaire. Simple percentage analysis, chi-square analysis, and correlation statistical tools were applied. The study found that salary, job autonomy, and recognition are important factors influencing employee job satisfaction. Fair compensation and freedom at work strongly affect how employees feel about their jobs. The study concludes that job satisfaction depends on multiple workplace factors working together, and that higher job satisfaction leads to better productivity and employee retention.

Licence: creative commons attribution 4.0

  License

Creative Commons Attribution 4.0 and The Open Definition

 Keywords

Job Satisfaction, Employee Productivity, Salary and Compensation, Job Autonomy, Employee Recognition, Organizational Performance

  License

Creative Commons Attribution 4.0 and The Open Definition

 Abstract

The convergence of Internet of Things (IoT), Machine Learning (ML), and Deep Learning (DL) has transformed traditional agriculture into a data-driven, precise, and sustainable paradigm known as Smart Agriculture 4.0. This survey systematically reviews 278 peer-reviewed studies from 2015 to 2025, categorizing applications into seven domains: crop health monitoring, yield prediction, water and soil management, livestock monitoring, greenhouse automation, supply-chain traceability, and agricultural drones. We propose a three-layer IoT-ML-DL taxonomy mapping sensing, feature extraction, and decision-making. Analysis covers 42 ML/DL architectures, 19 protocols, and 35 datasets. CNNs dominate imagery tasks with significant accuracy gains, while LSTMs excel in yield forecasting. Lightweight models enable edge deployment. Challenges include data heterogeneity, energy limits, connectivity in rural areas, and lack of benchmarks, particularly in developing nations. We suggest future directions like federated learning, digital twins, and 6G integration. This reference aids researchers and policymakers in advancing intelligent systems for food security.

Licence: creative commons attribution 4.0

  License

Creative Commons Attribution 4.0 and The Open Definition

 Keywords

Smart Agriculture, Internet of Things, Machine Learning, Deep Learning, Precision Farming, Edge AI, Digital Twin, Federated Learning

  License

Creative Commons Attribution 4.0 and The Open Definition

 Abstract

Changes in the demands of education require innovation and creativity in the learning process. With the development of Artificial Intelligence (AI) in the field of education to help process daily activities including teaching and learning. The objective of this study is to investigate Artificial Intelligence (AI) in education, especially in the teaching and learning process. Artificial Intelligence (AI) is the process of modeling human thinking and designing a machine so it can behave like humans. In the future progress of science and technology, teachers' work such as correcting, student attendance, giving daily tests and exams, explaining knowledge, making administrative reports and other systemic work can be submitted to be completed by technology devices. Teachers can save more energy and can focus more on non- systemic work to create a golden generation with more character and quality with natural intelligence where robots cannot do it. Technology only runs systemically and is automated based on human commands, while the human mind, especially teachers deliver new knowledge. Therefore, the teacher's intelligence will be unmatched. AI that emerged as the industrial revolution is also the result of the creative minds of human natural intelligence. So when compared, between the two will never have an equal position

Licence: creative commons attribution 4.0

  License

Creative Commons Attribution 4.0 and The Open Definition

 Keywords

Artificial Intelligence (AI), education, teaching and learning, History of Artificial Intelligence, Types of Artificial Intelligence, AI in Action: Core Components of Ai-Based Education.

  License

Creative Commons Attribution 4.0 and The Open Definition