Journal of Computer Based Parallel Programming

When AI Refactors Code: An Expert-informed Framework for Assessing Long-term Maintainability and Architectural Drift

2026-07-31T06:47:31+00:00

Artificial Intelligence (AI), including large language models, is increasingly used to identify, recommend, and implement software refactoring. Existing evaluations primarily emphasize immediate structural outcomes, including reduced complexity, improved cohesion, increased testability, and removal of code smells. However, limited guidance exists for determining whether such improvements remain beneficial across multiple development cycles or contribute to architectural drift, technical debt, and knowledge loss. This study develops an expert-informed framework for assessing long-term maintainability in AI-refactored software systems. A focused narrative synthesis of literature on software refactoring, maintainability, architectural erosion, software evolution, and AI-assisted software engineering informed a preliminary conceptual model. The model was subsequently refined through semi-structured interviews with three senior practitioners representing quality engineering, enterprise architecture, cloud infrastructure, and production operations. Hybrid deductive-inductive thematic analysis produced four integrated findings: long-term maintainability is multidimensional and temporal; repeated context-limited optimization may accumulate into architectural drift; assessment requires structural, architectural, contextual, governance, and operational evidence; and expert knowledge should be embedded through persistent architectural context, machine-enforceable controls, risk-based review, and operational feedback. These findings informed the AI Refactoring Maintainability and Drift Assessment Framework (AI-RMDAF), which links an approved system baseline, multidimensional assessment, risk classification, deployment monitoring, and continuous organizational learning. The study contributes an exploratory conceptual framework that provides a foundation for future empirical and longitudinal validation rather than a statistically or predictively validated model. It offers researchers a structured agenda for longitudinal evaluation and provides practitioners with a governance-oriented approach for preserving architectural intent while retaining the productivity advantages of AI-assisted refactoring.

AI for Sustainable Innovation: Translating Insights into Strategic Action

2026-07-21T04:34:27+00:00

Artificial Intelligence (AI) has great potential to promote sustainable innovation. It can help tackle urgent environmental, social, and economic issues by improving resource use, forecasting, and integrating systems across various sectors. However, despite the rising interest, organizations find it challenging to turn AI insights into practical sustainability actions. This struggle stems from fragmented structures, poor governance, and complicated ethical issues. This research aims to close these gaps by creating and testing a clear framework that explains how to use AI to support sustainable innovation while meeting sustainability goals. The study uses a mixed-methods approach, combining a systematic literature review with semi-structured interviews of experts from academia, industry, and policy. This method allows for a detailed summary of various knowledge areas and highlights real-world challenges and best practices for applying AI to sustainability. The findings identify seven key aspects. These include AI's role in integrating systems, gaps in turning insights into action, issues with data fragmentation, governance problems, ethical and equity concerns, sustainability challenges posed by AI, and the need for responsible, context-aware use of AI. The validated framework gives strong theoretical and practical support for organizations and policymakers as they explore AI's potential for sustainable innovation. It highlights the need to align technological opportunities with organizational readiness, ethical governance, and strategic decision-making. This work enriches academic discussions by blending perspectives from technology, management, ethics, and policy. It also provides practical advice to promote resilient and fair innovation systems in the changing landscape of digital sustainability.

AquaClean AI: Intelligent Underwater Trash Detection using Hybrid CNN and YOLO-based Deep Learning for Marine Pollution Monitoring

2026-06-25T11:53:46+00:00

Marine pollution is emerging as a critical global issue that threatens aquatic ecosystems and biodiversity at multiple levels. The accumulation of underwater waste disrupts ecological balance and demands intelligent monitoring solutions. Within the paradigm of deep learning-driven environmental systems, AquaClean AI is conceptualized as an intelligent framework that enables automated detection of underwater trash. The system integrates hybrid convolutional neural networks combining VGG16 and EfficientNet to extract deep visual representations from complex underwater imagery. It further employs YOLO-based object detection to localize and classify debris under challenging environmental conditions. The framework operationalizes computer vision principles to process distorted images affected by low visibility and noise while ensuring reliable detection performance. By synthesizing feature extraction and real-time detection into a unified pipeline, the system transforms raw underwater data into meaningful insights. The proposed approach demonstrates strong adaptability across diverse underwater scenarios and supports continuous monitoring applications. Experimental evaluation demonstrates high detection accuracy, with YOLO-based models achieving mAP of 0.96, precision of 0.94-0.96, and recall of 0.91-0.93, outperforming traditional detection approaches. The solution offers a scalable, intelligent approach to addressing underwater pollution using advanced artificial intelligence techniques.

A Comprehensive Review of Pre-Processing Methods for Legal Document Mining and Analysis

2026-07-27T10:47:41+00:00

Legal documents such as court judgments, statutes, contracts, and compliance reports are often lengthy, complex, and difficult to interpret manually due to their formal language, archaic terminology, and domain-specific structure. The increasing volume of digital legal content, driven by the digitization of judicial and regulatory systems, has created a growing need for automated systems capable of efficiently processing, analyzing, and summarizing legal text while preserving semantic meaning and contextual relevance. This paper presents a comprehensive AI-enabled framework for legal document mining and analysis with a strong emphasis on pre-processing techniques and thematic interpretation. The proposed methodology integrates Natural Language Processing (NLP)-based pre-processing operations such as OCR correction, tokenization, stopword removal, punctuation filtering, case normalization, and lemmatization to produce clean and semantically meaningful legal text, while avoiding the over-aggressive cleaning that can degrade legal-specific terms and citations. Transformer-based embeddings such as BERT and Legal-BERT are then employed to capture contextual and doctrinal semantics from the processed corpus. To address the multi-dimensional nature of legal documents, fuzzy topic modeling is utilized to identify overlapping legal themes with varying degrees of membership, such as jurisdiction, liability, and statutory interpretation. The framework further applies extractive summarization using ranking and redundancy-control strategies to generate concise summaries without losing critical legal information. In addition, thematic mapping techniques are used to visually represent relationships among statutes, legal arguments, provisions, and precedents. Experimental observations and comparative evaluation using ROUGE-based metrics demonstrate that the proposed approach achieves strong summarization quality, improved interpretability, and scalable performance for legal analytics applications.

MediCheck: A Design Thinking Approach to Early-Stage Childhood Cancer Detection and Geolocation-Enabled Diagnostic Support in India

2026-06-26T12:12:09+00:00

Catching childhood cancer early is the single most critical factor in determining survival. In India, where roughly 75,000 children are diagnosed every year, survival rates remain low, mostly because of delays in finding care, rural-urban disparities, and a lack of baseline awareness. In this study, they share the design and implementation of MediCheck, a web-based clinical screening and hospital mapping application built to guide parents from initial symptom concern to professional oncology consultations. Following a five-stage Design Thinking process (Empathize, Define, Ideate, Prototype, and Test), they structured the tool around a weighted symptom overlap index, real-time explainability features, and side-by-side benign differential diagnosis cards to ease parent anxiety. Geolocation-aware mapping connects users directly with oncology units via the Google Places API. To check the system's underlying logic, they ran two validation pipelines: a clinical risk classifier trained on synthetic symptom records (achieving 98.9% accuracy with an RF-Gradient Boosting ensemble) and a cell smear classifier trained on augmented blood images (reaching 95% accuracy in flagging lymphoblastic leukemia). User feedback showed that the platform helps parents organize symptom timelines, making consultations with doctors more effective.