Search

The pulp and paper industry is critical to global industrial and economic development. Recently, India's pulp and paper industries have been facing severe competitive challenges. The challenges have impaired the environmental performance and resulted in the closure of several operations. Assessment and prediction of the performance of the Indian pulp and paper industry using various parameters is a critical task for researchers. This study proposes a framework for performance assessment and prediction based on Data Envelopment Analysis (DEA), Artificial Neural Networks, and Deep Learning (DL) to assist industry administration and decision-making. We presented a case study based on eight industries to demonstrate the methodology's applicability. This study analyses and predicts industry performance based on sample data observations over 30 years. The result suggests the DEA-DL-based efficiency prediction has an overall MSE of 0.08 compared with the actual efficiency. Furthermore, the efficiency rankings are compared between the three techniques. The results suggest that the integrated DEA-DL method is primarily accurate in most scenarios with the actual values. The findings of this study provide a comprehensive analysis of environmental performance for policymakers. [ABSTRACT FROM AUTHOR]

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 text discusses the intersection of natural language understanding (NLU) and reasoning in the context of large language models (LLMs) and traditional logic-based approaches. It highlights the strengths and weaknesses of both approaches and explores the potential for hybrid models that combine symbolic and distributional representations. The text also mentions specific applications of hybrid approaches in natural language inference, question-answering, sentiment analysis, and dialog. The document concludes by introducing a special issue that features selected contributions from the NALOMA workshop series, which focuses on hybrid methods in NLU. [Extracted from the article]

Paper classification plays a pivotal role in facilitating precise literature retrieval, recommendations, and bibliometric analyses. However, current text-based methods predominantly emphasize intrinsic features such as titles, abstracts, and keywords, overlooking the valuable insights concealed within reference papers (i.e., cited papers). As a result, this oversight leads to reduced classification accuracy. In contrast, as a practical deep learning approach, graph neural networks incorporate the characteristics of reference papers to enhance paper classification. Nevertheless, traditional graph neural networks encounter limitations when handling intricate multi-level citation relationships in academic papers. To address these challenges, we introduce an enhanced graph neural network model for academic paper classification. This model integrates a multi-head attention mechanism and a residual network structure to dynamically allocate weights to various nodes within the graph, thereby enhancing its ability to handle complex multi-level citation relationships. Our experimental findings on an extensive real-world dataset demonstrate that our model achieves an accuracy of 61%, surpassing traditional graph neural networks by over 4%. Additionally, we have made the relevant datasets and models accessible on our GitHub repository. (https://github.com/xuejianhuang/ResGAT-for-paper-classification). [ABSTRACT FROM AUTHOR]

Citations are crucial artifacts to provide additional information to the reader to comprehend the research under concern. There are different roles that citations play in scientific discourse. Correctly identifying the intent of the citations finds applications ranging from predicting scholarly impact, finding idea propagation, to text summarization. With the rapid growth in scientific literature, the need for automated methods to classify citations is now growing intense. However, we can only fully understand the intent of a citation if we look at the citation context in the citing paper and also the primary purpose of the cited article. In this work, we propose a neural multi-task learning framework that harnesses the structural information of the research papers and the cited paper's information for the effective classification of citation intents. We analyze the impact of three auxiliary tasks on the performance of our approach for citation classification. Our experiments on three benchmark citation classification datasets show that incorporating cited paper information (title) shows that our deep neural model achieves a new state-of-the-art on the ACL-ARC dataset with an absolute increase of 5.3% in the F1 score over the previous best model. We also achieve comparable performance with respect to the best-performing systems in the SDP 2021 3C Shared task on Citation Context Classification. We make our codes available at https://github.com/Tirthankar-Ghosal/citationclassification-SCIM [ABSTRACT FROM AUTHOR]

With the global energy shortage, countries all over the world are vigorously developing new energy sources, and photovoltaic glass, as an important raw material for photovoltaic power generation, puts forward higher requirements for its output and quality. In order to solve the problems of low efficiency, susceptibility to interference by human factors, and low detection accuracy during the detection of photovoltaic glass edge defects by traditional manual methods, this paper proposes an automatic detection method of photovoltaic glass edge defects based on machine vision technology. Firstly, a machine vision defect detection system is designed to meet the needs of photovoltaic glass edge defect detection, includes high-contrast imaging solutions with a combination of multiple light sources for illumination and an automated transmission scheme with high stability, and an image dataset is established; Secondly, according to the characteristics of defect detection process with many interfering factors and high requirements for detection efficiency and accuracy, a deep learning defect detection method is proposed to improve the SqueezeNet model, which incorporates dense residual units into the two-part Fire Module of the classical SqueezeNet network model to extract the important feature information of the glass edge image, and effectively avoids the influence of interfering factors, such as water droplets, on the detection of defects. Finally, the improved SqueezeNet network algorithm is applied in the machine vision glass edge defect detection system designed in this paper. The experimental results show that the average leakage rate of the photovoltaic glass edge defect detection method proposed in this paper is 0.0064%, the misdetection detection rate is 0.0075%, and the average detection time is 2.715 s, and can meet the requirements of the automated production of photovoltaic glass. [ABSTRACT FROM AUTHOR]

Weld seam detection is an important part of automated welding. At present, few studies have been conducted on annular weld seams, and a lot of defects exist in the point cloud model of the tube sheet obtained by RGB-D cameras and photography methods. Aiming at the above problems, this paper proposed an annular weld seam detection network named WeldNet where a voxel feature encoding layer was adaptively improved for annular weld seams, the sparse convolutional network and region proposal network (RPN) were used to detect annular weld seam position, and an annular weld seam detection loss function was designed. Further, an annular weld seam dataset was established to train the network. Compared with the random sampling consistency (RANSAC) method, WeldNet has a higher detection accuracy, as well as a higher detection success rate which has increased by 23%. Compared with U-Net, WeldNet has been proven to achieve a better detection result, and the intersection over the union of the weld seam detection is improved by 17.8%. [ABSTRACT FROM AUTHOR]

Logistics parcel detection technology is critical for unmanned sorting. YOLOv5, the state-of-the-art (SOTA) object detection model, is a classic network widely used in engineering. However, on the premise of fast and accurate detection of logistics parcels, it confronts certain challenges of high computing load and parameter quantity. To address these issues, this paper proposes a two-scale lightweight deep learning model named SFYOLOv5. In this article, a lightweight feature extraction module called Pruned-Shuffle-Block (PSB) is proposed. Meanwhile, a double-layer pyramid structure for the medium and large target detection is created in accordance with the image size distribution of logistics parcels. With this structure, the Floating Point of Operations (FLOPs) and parameters in feature extraction were significantly reduced. In addition, a down sampling module named Focus for Downsampling (FFD) is designed and attention modules are introduced to extract high-level semantic information in logistics parcels. These modules not only compensate for the loss caused by down sampling but also improve the mean Average Precision (mAP). Finally, the comparison experiment is performed by using the self-built logistics parcel dataset. The results show that the mAP of the model reaches 99.1%, the number of model parameters decreased by 92.14%, and the FLOPs decreased by 89.57% compared with the existing SOTA model. This model can be used in logistics parcel intelligent sorting. [ABSTRACT FROM AUTHOR]

Traffic sign recognition, as one of the key steps of intelligent driving technologies, effectively avoids most traffic accidents by detecting the location and type of traffic signs in real time and providing the information to drivers or autonomous vehicles promptly. In addition, edge devices close to users has become an inevitable requirement for the development of IoT technology, real-time computing and the realization of network edge intelligence. Nowadays, the YOLO algorithm in object detection, developed to YOLOv8, is accompanied by various lightweight networks and lightweight methods to win by "fast", so this paper will propose an algorithm, fusing Ghost module and Efficient Multi-Scale Attention Module into YOLOv8, so that the model can improve the computing speed while maintaining the original characteristics. Furthermore, we choose Raspberry Pi as the object detection device due to its many characteristics such as lightweight, low power consumption. Through experiments, the model is trained on CCTSDB dataset, and the improved algorithm is tested on Raspberry Pi 4B. The results show that for three types of traffic signs, namely prohibited, indication and warning, the recognition accuracy mAP reaches 93.5% on the poor weather test set and 82.9% on the original test set, and the inference delay of Raspberry Pi reaches 0.79 s, which is effective in actual road test experiments. The improved model's accuracy has increased by 6.4% and 3.8% on two separate test sets compared to the original model, while the detection time has been reduced by 0.12 s. This research is of great significance to the maturation of assisted driving and autonomous driving technologies. [ABSTRACT FROM AUTHOR]

Recently, image inpainting techniques tend to be more concerned with how to enhance the quality of restoration than with how to function on various platforms with limited processing power. In this paper, we propose a lightweight method that combines group convolution and attention mechanism to improve or replace the traditional convolution module. Group convolution was used to achieve multi-level image inpainting, and the authors proposed the rotating attention mechanism for allocation to deal with the issue of information mobility between channels in traditional convolution processing. The parallel discriminator structure was utilized throughout the network's overall design phase to guarantee both local and global consistency of the image inpainting process. The experimental results can demonstrate that, while the quality of image inpainting has been ensured, the proposed image inpainting network's inference time and resource usage are significantly lower than those of comparable lightweight approaches. [ABSTRACT FROM AUTHOR]