Hasan Ali Khattak

Given the rapid increase of textual data in various fields, text summarization has become essential for efficient information handling. Over recent decades, numerous methods have been proposed to enhance summarization processes, and various review papers and books have been published to encapsulate these methodologies and discuss their implications. However, existing reviews often fail to provide a comprehensive retrospective of recent advancements, particularly concerning detailed architectural frameworks, the field’s current state, evaluation methodologies, and unresolved challenges. This paper addresses this gap by presenting a detailed analysis of the extractive approach, encompassing its inherent strengths, limitations, and underlying mechanisms. We present a detailed, multi-layered architectural framework designed to advance and develop summarization models, thereby supporting researchers in their endeavors. Furthermore, this review categorizes domain-specific summarization techniques, focusing on unique challenges and tailored strategies for news, scientific articles, and social media. We emphasize the impact of evaluation metrics and benchmark datasets in performance assessment, providing a detailed analysis of the commonly utilized datasets and metrics in the current literature. This paper is a valuable resource for advancing text summarization techniques in natural language processing and machine learning by identifying future research directions and open challenges.

Unmanned Aerial Vehicles (UAVs) have revolutionized various industries by providing efficient and automated flight capabilities. However, reliance on GPS and traditional navigation systems poses challenges in scenarios where signal interference or failures occur. In this research, we present a novel computer vision-based method to enhance UAV navigation, enabling accurate height and location estimation. Our approach utilizes a sophisticated network that leverages a pair of images to estimate UAV height. The Pyramid Stereo Matching Network is employed to extract robust image features and generate a disparity map. Subsequently, a custom network processes and convolves this data, employing diverse computer vision techniques to achieve precise height estimation. To evaluate the effectiveness of our proposed method, we collected a comprehensive dataset by conducting flights with a Phantom 4 Pro drone over NUST Main campus, H-12 Islamabad. The dataset encompasses images captured at ten different heights, spanning from 100 to 280 meters, with flights evenly spaced 20 meters apart. In rigorous evaluations, our approach demonstrates promising results compared to existing methods. By liberating UAVs from reliance on GPS, this vision-based 3D localization technique holds immense potential to ensure successful flights even in challenging environments.

The Holy Quran has the utmost importance for the Muslim community, and to get a full reward, the Quran should be read according to the rules mentioned. In the past few years, this field has gained a lot of importance in the eyes of researchers who aim to automate the Quranic reading and understanding process with the help of Machine Learning and Deep Learning, knowing it has a lot of challenges. To date, there are a lot of research categories explored. However, still, there lacks a few holistic, including one detailed survey of all the categories and methodologies used to solve problems. We focused the paper on being a one-stop-shop for the people interested so they could find (i) all related information and (ii) future gaps in research. This paper provides a detailed survey on Deep Modeling for Quranic Recitation to address these challenges. We discussed all possible categories of speech analysis, including the most advanced feature extraction techniques, mispronunciation detection using Tajweed rules, Reciters and speech dialect classification, and implementation of Automatic Speech Recognition (ASR) on Quranic Recitations. We also discussed research challenges in this domain and identified possible future gaps.

Machine learning provides a powerful mechanism to enhance the capabilities of the next generation of smart cities. Whether healthcare monitoring, building automation, energy management, or traffic management, use cases of capability enhancement using machine learning have been significant in recent years. This paper proposes a modeling approach for scheduling energy consumption within smart homes based on a non-dominated sorting genetic algorithm (NSGA). Distributed energy management plays a significant role in reducing energy consumption and carbon emissions as compared to centralized energy generation. Multiple energy consumers can schedule energy-consuming household tasks using home energy management systems in coordination to reduce economic costs and greenhouse gas emissions. In this work, such a home energy management system is used to collect energy price data from the electricity company via an embedded device-enabled smart meter and schedule energy consumption tasks based on this data. We schedule daily power consumption tasks using a multiobjective optimization method that considers environmental and economic sustainability. Two conflicting objectives are minimizing daily energy costs and reducing carbon dioxide emissions. Based on electricity tariffs, CO₂ intensity, and the window of time during which electricity is consumed, energy consumption tasks involving distributed energy resources (DERs) and electricity consumption are scheduled. The proposed model is implemented in a model smart building consisting of 30 homes under 3 pricing schemes. The energy demand is spread out across a 24-hour period for points A2–A4 under CPP-PDC, which produces a more flattened curve than point A1. There are competing goals between electricity costs and carbon footprints at points B2–B4 under the CPP-PDC, where electricity demand is set between 20:00 and 0:00. Power grids’ peak energy demand is comparatively low when scheduling under CPP-PDC for points A5 and B5. Reducing carbon emissions, CPP-PDC reduces the maximum demand for electricity from the grid and the overall demand above the predetermined level. The maximum power demand from the grid is minimized for points A5 and B5, reducing up to 22% compared to A2. The proposed method minimizes both energy costs as well as CO₂ emissions. A Pareto curve illustrates the trade-off between cost and CO₂ emissions.

Edge computing, a distributed computing architecture within the knowledge-defined network (KDN), faces challenges due to the significant disparities and data heterogeneity among its nodes, hindering their interaction. Ontology, a solution within the Semantic Web, is well-suited for addressing data heterogeneity and matching ontologies effectively. However, ontology matching presents difficulties due to non-linear mathematical issues. To overcome these challenges, the generative adversarial network (GAN), an unsupervised learning method, has emerged as a promising tool. GAN consists of two models with distinct objectives trained against eachother to achieve optimal outcomes. This paper introduces SA-GAN, an algorithm that combines GAN with simulation-based annealing to enhance its effectiveness. SA-GAN utilizes a stagnation counter to expedite the convergence speed of GAN. Through experiments conducted on a renowned ontology benchmark, the paper demonstrates that SA-GAN, along with other ontology matching algorithms, can identify the best alignments. Consequently, SA-GAN facilitates the construction of bridges in edge computing, improving its overall effectiveness.

Software-defined networking (SDN) has introduced flexibility and ease of management in the network. SDN offers centralized and fine-grained control over the web it has become an industry trend and is adopted in various areas of communication and networking. Controllers in SDN play a pivotal role, and their performance is key to the stability and scalability of the network. This study evaluates different SDN controllers i.e., NOX, POX and RYU. The performance of the controllers is analyzed based on Quality of Service (QoS) parameters, including Flow Setup Latency, Round Trip Time (RTT), Initial Ping Delay (IPD), Jitter, Throughput, and TCP & UDP Bandwidth. Different-sized networks are simulated, by varying the number of switches in linear and tree topologies and hosts in a single topology, using Mininet SDN simulator. Our results show that the performance of RYU is very consistent and shows small variation in the performance with increasing network size and traffic loads as compared to NOX and POX.

Pandemic and natural disasters are growing more often, imposing even more pressure on life care services and users. There are knowledge gaps regarding how to prevent disasters and pandemics. In recent years, after heart disease, corona virus disease‐19 (COVID‐19), brain stroke, and cancer are at their peak. Different machine learning and deep learning‐based techniques are presented to detect these diseases. Existing technique uses two branches that have been used for detection and prediction of disease accurately such as brain hemorrhage. However, existing techniques have been focused on the detection of specific diseases with double‐branches convolutional neural networks (CNNs). There is a need to develop a model to detect multiple diseases at the same time using computerized tomography (CT) scan images. We proposed a model that consists of 12 branches of CNN to detect the different types of diseases with their subtypes using CT scan images and classify them more accurately. We proposed multi‐branch sustainable CNN model with deep learning architecture trained on the brain CT hemorrhage, COVID‐19 lung CT scans and chest CT scans with subtypes of lung cancers. Feature extracted automatically from preprocessed input data and passed to classifiers for classification in the form of concatenated feature vectors. Six classifiers support vector machine (SVM), decision tree (DT), K‐nearest neighbor (K‐NN), artificial neural network (ANN), naïve Bayes (NB), linear regression (LR) classifiers, and three ensembles the random forest (RF), AdaBoost, gradient boosting ensembles were tested on our model for classification and prediction. Our model achieved the best results on RF on each dataset. Respectively, on brain CT hemorrhage achieved (99.79%) accuracy, on COVID‐19 lung CT scans achieved (97.61%), and on chest CT scans dataset achieved (98.77%).