Search

Thousands of papers are published at conferences every year. Conference organizers manually assign accepted papers to the sessions according to the title and keywords given by the author. Then organizer names each session, based on his/her experience. These are complex and time-consuming processes and often fail to collect papers on a similar topic in content. This often causes the participant to exit the session after listening to the presentation of one or two papers, because the session name does not fully represent the papers in the session and the papers in the session are not close in content. As a solution to these problems, this paper proposes a method for automatically organizing conference sessions. The method first introduces a network created with a deep learning-based document similarity. Then, sessions are determined with a community discovery method specific to this network, and finally, session titles are extracted with a topic modeling approach. To the best of our knowledge, this paper is the first effort in this direction. Experiments conducted on sessions of three real conferences show that the proposed method is able to create up to 21% better similar sessions, and session names better represent the papers in that session. [ABSTRACT FROM AUTHOR]

The term "artificial intelligence" (AI) refers to "smart" high-tech that is mindful of and able to learn from its surroundings. It is the most revolutionary technology that humans have ever created. Common AI approaches involving machine learning and deep learning techniques can be effectively applied to resolve today's various cybersecurity issues. Furthermore, the metaverse is all about how people communicate and engage with one another through technology. This survey explores the role of AI with its emerging applications and their various technologies, such as the metaverse, healthcare, IoT, gaming, and many more. To determine the strengths, flaws, opportunities, and risks that are inherent in artificial intelligence technologies, using an extensive literature survey, the SWOT (Strengths, Weaknesses, Opportunities, and Threats) assessments have been undertaken in this survey paper. Finally, the survey paper summarises the current state of knowledge of AI applications and discusses the findings present in recent research to ensure a favourable change in artificial intelligence advances and applications. Some technical AI challenges, like high-speed, high-performance hardware and reducing the amount of training data, etc., are also discussed with future scope. [ABSTRACT FROM AUTHOR]

Distinguishing a natural scene from artwork is a simple task for the human visual system, but a challenging one for machines due to the wide range of psychovisual features, illumination gamuts, and varied interpretations of glossiness. While state-of-the-art image quality metrics quantify overall visual quality to a remarkable degree of accuracy, quantification of 'glossiness' of a scene to represent its naturalness is still an emerging area. The rapid growth of deep learning methods and CNN-based architectures inspired us to explore the fusion of best performing CNN architectures in this paper for an image dataset specifically created to replicate unnaturalness of artwork or portrait-like images. Performance of four CNN networks was tested using transfer learning, selective retraining and optimizing initial learning rates on a dataset of about 8.5k images created to represent various degrees of glossiness. A fusion framework was then proposed using the top two architectures. In terms of eleven levels of naturalness (0 to 10), both quantitative and qualitative evaluation of the fusion frameworks was conducted. The framework resulting from fusion of GoogleNet and VGG16, referred to as GoogleVGG Fusion in this paper, is found to reach accuracies comparable to individual networks but with nearly half the computational cost. The proposed GoogleVGG Fusion model achieved an accuracy of 87.86% with the labelled scores and a Spearman's Rank Correlation (SROCC) of 0.9794. As expected, the accuracy of the proposed framework with the subjective scores in comparison with non–deep learning (DL) & DL-based methods is remarkably better. [ABSTRACT FROM AUTHOR]

Aspect-base sentiment analysis (ABSA) refers to the analysis of sentiments in users' notes from the perspective of different aspects, which has attracted much attention in the last decade. With the development of social networks, user-generated content on various Internet platforms is growing rapidly. This content contains valuable information, especially in e-commerce, which helps people and consumers make better decisions. Therefore, improving the performance of ABSA methods has recently attracted the attention of many researchers. In this paper, the basic concepts in sentiment analysis, applications, challenges, evaluation metrics, common benchmark datasets, required pre-processing and textual representation methods have been defined. An overview of all available methods for ABSA is also provided, and then a detailed review of deep learning methods in ABSA with a focus on graph neural networks has been done. Finally, this paper reviews the latest studies that have used deep learning and graph neural networks to solve sentiment analysis problems. We hope that our summaries in this paper can provide necessary guidelines for beginners and new researchers. [ABSTRACT FROM AUTHOR]

Noising in X-ray imaging has been one of the biggest challenges that leads to insufficient and improper diagnosis. Despite the fact that X-rays are one of the most widespread and acceptable imaging techniques among the medical and scientific fraternity, still Gaussian and Poisson noise lead to a lot of image deterioration. Over the past few decades, several denoising techniques have been explored using traditional, hybrid and deep learning techniques which have been reported in this paper. Poisson noise was best removed by the application of bilateral filter with a maximum Peak Signal to Noise Ratio (PSNR) of 36.22 and for the removal of Gaussian noise, median filter proved to be unparalleled with a PSNR of 32.92 for the variance of 0.01, 31.4 for the variance of 0.04, 31.03 for the variance of 0.07, and 30.58 for the variance of 0.1 amongst the conventional filters. The Noise2Noise model employing the deep learning approach has given the best PSNR value of 34.38 amongst all the other alternatives for the images with gaussian noise. This paper serves as a comprehensive review for beginners working in this domain, that would aid them to select the best filter for the image pre-processing and noise removal. [ABSTRACT FROM AUTHOR]

The honking events' density reflects the level of traffic noise pollution, road congestion, etc in the urban areas. In this paper, we propose a participatory sensing based traffic honk monitoring system called HonkSense that uses smartphone equipped sensors (e.g. microphone, GPS, etc.). Citizens can take part in monitoring traffic noise pollution due to honking by recording ambient noise on the road. Application running on users' smartphones is used to extract features in real time from recorded audio and then send to the cloud for honk detection and decision making tasks. Here, Mel-Frequency Cepstral Coefficients (MFCCs) are utilized as feature for presenting audio signals in honk detection. This paper uses a deep Convolutional Neural Network (CNN) model that is deployed to cloud for detecting traffic honking events. The end-to-end system provides a privacy-preserving (anonymous data collection), low-power and low-cost solution for participatory sensing based traffic honk monitoring. We evaluate our proposed system on real world participatory sensing based road sound dataset collected by participants. It achieves a classification accuracy of 96.3%. The deep CNN is also evaluated on different benchmark datasets (namely ESC-50 and UrbanSound8K). The results are also compared with the baseline support vector machine (SVM) and k-nearest neighbors (KNN) classification models. Besides, state-of-the-art visualization techniques are used to explore spatial and temporal variability of honking events in urban areas using two case studies. [ABSTRACT FROM AUTHOR]

Social media sites are now quite popular among internet users for sharing news and opinions. This has become possible due to the inexpensive Internet, easy availability of digital devices, and the no-cost policy to create a user account on social media platforms. People are fascinated by social media sites because they can easily connect with others to share their interests, news, and opinions. According to studies, someone who lacks credibility is more likely to spread false information in order to achieve goals of any kind, be it influencing political opinions, earning attention, or making money. The automatic detection of social media related fake news has thus emerged as a highly anticipated research area in recent years. This paper offers a comprehensive review of the automatic detection of fake news on social media platforms. It contains details of the key models or techniques related to machine/deep learning proposed (or developed) during the period of the year 2011 to the year 2022 along with the performance metrics of each model or technique. The paper discusses (a). the key challenges faced during the development of an effective and efficient fake news detection system, (b). some popular datasets for carrying out fake news detection related research, and (c). the major research gaps, and future research directions in the area of automatic fake news detection on social media. [ABSTRACT FROM AUTHOR]

Image Captioning is the process of generating textual descriptions of given images. It encompasses two major fields of deep learning, computer vision, and natural language processing. This paper presents an Image Captioning model which uses the Convolution Neural Network (CNN) model for feature extraction and a transformer architecture for the generation of sequences from these feature vectors. For feature extraction, this paper uses different CNN architectures like Xception, InceptionV3, ResNet50V2, VGG19, DenseNet201, ResNet152V2, EfficientNetV2B3, EfficientNetV2B0. The proposed method takes advantage of the transformer model for faster processing, and Beam search is implemented to get the top N most probable sequences for each image. The architecture is trained on Flickr8k dataset and the model outperforms the existing methods. The proposed model achieves a BLEU_4 score of 0.2184 on the Flickr8k dataset. [ABSTRACT FROM AUTHOR]

Deep neural networks (DNNs) have achieved great success in many object detection tasks. However, such DNNS-based large object detection models are generally computationally expensive and memory intensive. It is difficult to deploy them to devices with low memory resources or scenarios with high real-time requirements, which greatly limits their application and promotion. In recent years, many researchers have focused on compressing large object detection models without significantly degrading their performance, and have made great progress. Therefore, this paper presents a survey of object detection model compression techniques in recent years. Firstly, these compression techniques were divided into six categories: network pruning, lightweight network design, neural architecture search (NAS), low-rank decomposition, network quantization, and Knowledge distillation (KD) methods. For each category, we select some representative state-of-the-art methods and compare and analyze their performance on public datasets. After that, we discuss the application scenarios and future directions of model compression techniques. Finally, this paper is further concluded by analyzing the advantages and disadvantages of six types of model compression techniques. [ABSTRACT FROM AUTHOR]

Predicting student performance is a critical aspect of educational systems. Although forecasting a student's future performance is essential in many applications, it is a challenging process due to various factors. Previous research in this area has mainly focused on comparing machine learning methods to automate student evaluation and predict their final performance. However, there have been limited studies that thoroughly explore the issue of class imbalance using a deep learning approach. Moreover, the large dataset targeting university students makes it well-suited for in-depth analysis and increases the likelihood of obtaining more accurate results. This study presents a deep learning model based on convolution and introduces a comprehensive exploration of oversampling and undersampling methods to address the issue of imbalanced classes. The paper investigates various features and characteristics of undergraduate students at the University of Jordan, utilizing a large dataset collected from the university's registration unit. These features include demographic information, attributes related to students' majors, faculties, registrations, courses taken (such as passed, repeated, and completed), as well as their high school averages and performance in the first four semesters. The results demonstrate that the model performs exceptionally well in terms of gmean when predicting students' excellence. This research project has significant implications and provides valuable insights to the research community and higher education managers, aiding in the development of improved strategies to enhance educational performance. Future researchers can utilize the methods employed in this paper during the data preprocessing stages and implement the demonstrated balancing strategies for further advancements in this field of study. [ABSTRACT FROM AUTHOR]