Volume 14 | Issue 6 |

 Abstract

Integrating multiple APIs, third-party services, and AI models into single comprehensive workflow has become more difficult in this increasing growth of modern applications. Traditional workflow automation platforms often struggle to manage such complexity, especially when working with long and inter-related tasks, real-time data processing and AI-driven operations. This platform frequently faces issues such as API timeouts, inefficient resource utilization and loss of data consistency between different stages of workflow. Challenges such as maintaining context across multiple steps and ensuring reliability between interconnected components in the system often occurs while integrating different AI models and Services. These drawbacks make current systems difficult to scale and not suited for real-world production environment.

Licence: creative commons attribution 4.0

  License

Creative Commons Attribution 4.0 and The Open Definition

 Keywords

AI Workflow Orchestration, Type-Safe Full-Stack Systems, Multi-Model LLM Integration, Asynchronous Task Execution, Distributed Workflow Systems, Context-Aware Data Flow

  License

Creative Commons Attribution 4.0 and The Open Definition

 Abstract

Cultural heritage sites often face challenges such as structural deterioration, environmental damage, or limited accessibility, making preservation and public engagement dif-ficult. Emerging digital technologies, particularly Augmented Reality (AR), provide innovative ways to experience and interact with such monuments. This paper presents an AR-based system designed to facilitate interactive heritage exploration through digital reconstruction, intelligent narration, and community en-gagement.

Licence: creative commons attribution 4.0

  License

Creative Commons Attribution 4.0 and The Open Definition

 Keywords

Augmented Reality, Cultural Heritage, Mon-ument Reconstruction, AI Assistant, ChatGPT API, Tourism, Education

  License

Creative Commons Attribution 4.0 and The Open Definition

 Abstract

The freshness of fruits is critical for maintaining quality in fruit. This paper represents an automated fruit freshness detection system using YOLOv8 object detection with squeeze-and-Excitation. the freshness of fruits our method balances accuracy and real time performance while detecting the freshness level of fruits the average accuracy is 85.5% with a maximum accuracy of 94.7% for detecting accuracy of 94.7% for detecting the freshness level of single fruits category .the results confirm that the enhanced YOLOv8 - SE model effectively detects and localizes fruit freshness conditions , indicating its suitability for smart agriculture and food quality monitoring applications.

Licence: creative commons attribution 4.0

  License

Creative Commons Attribution 4.0 and The Open Definition

 Keywords

Fruit recognition, freshness detection, image classification, MobileNet2, computer vision, K means clustering, deep learning, nutrition analysis, React.js, flask.

  License

Creative Commons Attribution 4.0 and The Open Definition

 Abstract

The fingertip acts as a virtual colored marker. Using Open CV, the captured frames are converted into HSV colour space, where colour detection and segmentation techniques are applied to isolate the fingertip region. The detected fingertip coordinates are continuously tracked and mapped onto a digital canvas, allowing the system to draw strokes corresponding to the user's hand movement. With designers and performers using digital media, touch and gesture-based interfaces are increasingly widely used in the creative industry. There are issues or flaws with these touchscreens, such as ergonomic strain and accuracy. The goal of this research is to thoroughly analyse how well these interfaces facilitate accuracy, expressiveness, and intuitive interaction while taking ergonomics into account.

Licence: creative commons attribution 4.0

  License

Creative Commons Attribution 4.0 and The Open Definition

 Keywords

computer vision, virtual drawing, human-computer interaction, gesture recognition, hand tracking, and real-time processing

  License

Creative Commons Attribution 4.0 and The Open Definition

 Abstract

This paper presents a critical review and comparative evaluation of machine learning methods to predict the compressive strength of concrete with particular emphasis on mixes that incorporate supplementary cementitious materials. The analysis includes ensemble algorithms (e.g., Random Forest, CatBoost, XGBoost, AdaBoost, Gradient Boosting), regression algorithms (Support Vector Regression, Linear Regression), and neural networks (Artificial Neural Networks, Extreme Learning Machines) by synthesizing the results of the latest studies. CatBoost was the most effective one, with the highest R 2 values of up to 0.94 in several studies. The most significant parameters found are concrete age, water-to-binder ratio, and cement content. Other important gaps in current literature that are identified during the review are inconsistency in datasets, lack of external model validation, lack of interest in hybrid models, and real-world implementation barriers. These results provide a basis on which universal and predictable forecasting models can be developed to more sustainably build concrete constructions.

Licence: creative commons attribution 4.0

  License

Creative Commons Attribution 4.0 and The Open Definition

 Keywords

RandomForest,,CatBoost,,LinearRegression,,NeuralNetwork

  License

Creative Commons Attribution 4.0 and The Open Definition

 Abstract

Word Sense Disambiguation (WSD) is a critical task in Natural Language Processing (NLP) that seeks to resolve lexical ambiguity by determining the appropriate meaning of a word within a specific context. While considerable progress has been made in WSD for high-resource languages, low-resource languages have received limited focus due to a lack of annotated datasets, lexical resources, and computational tools. This paper provides a thorough survey of current WSD techniques, including knowledge-based, supervised, unsupervised, and deep learning methods, and assesses their applicability in low-resource language environments. The study also addresses the challenges faced by low-resource languages, such as data scarcity, morphological complexity, the absence of standardized tools, and multilingual influences. Furthermore, it reviews WSD research in Indian languages, including Manipuri, Malayalam, Punjabi, Bengali, and Sindhi, to highlight both current advancements and limitations. Notably, Sindhi is recognized as one of the least investigated languages concerning WSD research. The findings of this survey underscore the necessity for developing resource-efficient and adaptable approaches, particularly leveraging machine learning and deep learning techniques, to enhance WSD performance in low-resource languages. The paper also outlines future research directions aimed at overcoming existing challenges and closing the research gap in this area.

Licence: creative commons attribution 4.0

  License

Creative Commons Attribution 4.0 and The Open Definition

 Keywords

Word Sense Disambiguation, Low-Resource Languages, NLP, Machine Learning, Deep Learning, Sindhi

  License

Creative Commons Attribution 4.0 and The Open Definition

 Abstract

Driver fatigue and drowsiness are significant contributors to road traffic accidents worldwide, posing serious threats to public safety and increasing the need for reliable driver monitoring technologies. Vision-based driver drowsiness detection systems have emerged as an effective and non-intrusive solution for monitoring driver behavior using computer vision and artificial intelligence techniques. These systems analyze visual indicators such as eye closure, blink rate, yawning behavior, head pose, and gaze direction to assess the driver's level of alertness. This paper presents a PRISMA-guided systematic review of vision-based driver drowsiness detection approaches, focusing on detection algorithms, fatigue indicators, machine learning techniques and real-time monitoring architectures. The review examines widely used techniques such as facial landmark detection, Eye Aspect Ratio (EAR), PERCLOS-based fatigue estimation, yawning detection, gaze tracking, and deep learning models. Comparative analysis of existing methods indicates that deep learning and hybrid vision-based approaches significantly enhance detection accuracy, although challenges related to illumination variation, occlusion, head pose changes, and real-time deployment remain. The study highlights current research trends and outlines future directions for developing robust and efficient driver monitoring systems for intelligent transportation environments.

Licence: creative commons attribution 4.0

  License

Creative Commons Attribution 4.0 and The Open Definition

 Keywords

Driver Drowsiness Detection, Computer Vision, Eye Aspect Ratio (EAR), PERCLOS, Deep Learning, Driver Monitoring Systems, Intelligent Transportation Systems, PRISMA Systematic Review.

  License

Creative Commons Attribution 4.0 and The Open Definition

 Abstract

Deepfake technologies driven by generative adversarial networks (GANs), neural vocoders, and end-to-end speech synthesis architectures have significantly advanced the generation of highly realistic synthetic audio. Modern voice cloning systems are capable of replicating speaker identity, linguistic style, and vocal timbre with minimal training data, making synthetic speech increasingly indistinguishable from authentic human recordings. While these developments offer substantial benefits in assistive technologies, entertainment, virtual agents, and content creation, they simultaneously introduce serious security, ethical, and societal risks. Malicious applications include financial fraud, identity impersonation, political misinformation, social engineering attacks, and erosion of public trust in digital media.

Licence: creative commons attribution 4.0

  License

Creative Commons Attribution 4.0 and The Open Definition

 Keywords

CE-14 Deepfake Detection: Assessment of Speech and Emotion-Based Forensic Analysis

  License

Creative Commons Attribution 4.0 and The Open Definition

 Abstract

The proposed system is based on e-certificate system in India's educational framework that leverages blockchain technology to address the widespread issue of certificate forgery. The inherent qualities of blockchain, including its immutability and transparency, create a solid foundation for enhancing the security and reliability of educational certifications. The operational framework is built around creating and storing an electronic file containing essential academic information in a dedicated database. Concurrently, the system generates a unique hash value for this electronic file, which serves as an exclusive identifier. This hash is securely embedded within a blockchain block, benefiting from the technology's resistance to tampering. To enable verification processes, both an inquiry string code and a QR code are linked to each certificate, encapsulating necessary information for authenticity verification. Users can initiate validation by scanning the QR code with a mobile device or entering the inquiry string on a specific website. The hash stored in the blockchain is then checked to confirm that the certificate has not been altered. This proposed solution significantly enhances the credibility of traditional paper certificates by implementing a dependable, transparent, and tamper-resistant verification process.

Licence: creative commons attribution 4.0

  License

Creative Commons Attribution 4.0 and The Open Definition

 Keywords

Blockchain, Digital Certificate, Hashing, E-Certificate, Certificate Verification, Transparency, QR Code, Secure Storage, Credential Authentication.

  License

Creative Commons Attribution 4.0 and The Open Definition

 Abstract

Evaluating subjective answers remains a major challenge in educational assessment due to the wide variation in students' writing styles, vocabulary, and expression of concepts. Traditional grading methods, such as manual evaluation or rule-based systems, are time-consuming, inconsistent, and unable to capture the true semantic meaning of answers. This paper presents a machine learning-based framework for automated subjective answer evaluation that leverages Natural Language Processing (NLP) techniques to assess student responses more effectively. The proposed system utilizes semantic embeddings and transformer-based architectures to analyze the contextual meaning of answers, enabling it to recognize paraphrased expressions and evaluate the relevance, completeness, and coherence of responses. The model is trained and validated using the ASAP-SAS dataset and further adapted to evaluate teacher-uploaded question papers. Standard evaluation metrics such as Quadratic Weighted Kappa (QWK), F1-score, and Pearson correlation are used to measure performance. The system aims to reduce manual effort, improve grading fairness, and provide timely feedback to students. Experimental outcomes demonstrate that machine learning can offer a scalable, consistent, and intelligent alternative to traditional grading, enhancing both the efficiency and reliability of academic assessments.

Licence: creative commons attribution 4.0

  License

Creative Commons Attribution 4.0 and The Open Definition

 Keywords

Subjective Answer Evaluation, Machine Learning, Natural Language Processing (NLP), Semantic Similarity, Transformer Models, BERT, Automated Grading, Educational Assessment, Deep Learning, Text Evaluation.

  License

Creative Commons Attribution 4.0 and The Open Definition