检索结果-内蒙古大学图书馆

Leveraging Unlabeled Corpus for Arabic Dialect Identification

computers, Materials & Continua 2025年第5期83卷 3471-3491页

作者： Mohammed Abdelmajeed Jiangbin Zheng Ahmed Murtadha Youcef Nafa Mohammed Abaker Muhammad Pervez Akhter School of Computer Science Northwestern Polytechnical UniversityXi'an710072China Department of Computer Science Applied CollegeKing Khalid UniversityMuhayil6311Saudi Arabia Computer Science Department National University of Modern LanguagesFaisalabad38000Pakistan

Arabic Dialect Identification(DID)is a task in Natural Language Processing(NLP)that involves determining the dialect of a given piece of text in *** state-of-the-art solutions for DID are built on various deep neural networks that commonly learn the representation of sentences in response to a given *** the effectiveness of these solutions,the performance heavily relies on the amount of labeled examples,which is labor-intensive to atain and may not be readily available in real-world *** alleviate the burden of labeling data,this paper introduces a novel solution that leverages unlabeled corpora to boost performance on the DID ***,we design an architecture that enables learning the shared information between labeled and unlabeled texts through a gradient reversal *** key idea is to penalize the model for learning source dataset specific features and thus enable it to capture common knowledge regardless of the ***,we evaluate the proposed solution on benchmark datasets for *** extensive experiments show that it performs signifcantly better,especially,with sparse labeled *** comparing our approach with existing Pre-trained Language Models(PLMs),we achieve a new state-of-the-art performance in the DID *** code will be available on GitHub upon the paper's acceptance.

关键词： Arabic dialect identification natural language processing bidirectional encoder representations from transformers pre-trained language models gradient reversal layer

来源：评论

学校读者我要写书评

暂无评论

Enhancing financial time series forecasting through topological data analysis

引用

Neural Computing and Applications 2025年第9期37卷 6527-6545页

作者： de Jesus, Luiz Carlos Fernández-Navarro, Francisco Carbonero-Ruz, Mariano Department of Quantitative Methods Universidad Loyola Andalucía Córdoba41704 Spain Department of Computer Languages and Computer Science Universidad de Málaga Málaga29010 Spain

Topological data analysis (TDA) is increasingly acknowledged within financial markets for its capacity to manage complexity and discern nuanced patterns and structures. It has been applied effectively to uncover intricate relationships and capture non-linear dependencies inherent in market data. This manuscript presents a groundbreaking study that delves into integrating features derived from TDA to improve the performance of forecasting models for univariate time series prediction. The research specifically examines whether incorporating features extracted from TDA-such as entropy, amplitude, and the number of points obtained from persistent diagrams can provide valuable supplementary information to the baseline forecasting model. Thus, the aim is to determine if these TDA-derived features can boost forecasting accuracy by offering additional insights that existing models might overlook. The N-BEATS model serves as the baseline forecasting model due to its robust generalization capabilities and flexibility in incorporating additional features into the model. The proposed methodology is compared against a univariate N-BEATS model without additional features and other strategies incorporating supplementary features such as temporal decomposition and time delay embeddings. The evaluation includes forecasting for six cryptocurrencies across four distinct time scenarios and four traditional financial instruments across two scenarios each, resulting in 32 datasets. The results obtained were promising, as the proposed method, N-BEATS+TDA, achieved the best results in mean performance and mean ranking for the three metrics considered (MAPE, MAE, and RMSE). Significant differences were observed with the rest of the proposed methods using a significance level of α=0.10, highlighting the effectiveness of integrating TDA features to enhance forecasting models. © The Author(s) 2024.

关键词： Prediction models

来源：评论

学校读者我要写书评

暂无评论

Magnification Independent Breast Cancer Analysis Using Vision Transformer

引用

Multimedia Tools and Applications 2025年第5期84卷 2029-2057页

作者： Tariq, Shehroz Raza, Rehan Sargano, Allah Bux Habib, Zulfiqar Department of Computer Science COMSATS University Islamabad Lahore Campus Pakistan Department of Computer Science National University of Modern Languages Islamabad Faisalabad Campus Pakistan School of Information Technology Murdoch University Perth Australia

Breast Cancer (BC) remains a critical global health challenge, accounting for millions of lives annually. Most breast cancer detection techniques are designed for binary classification. However, further classifying these types into their subtypes remains challenging due to cell coherence, variations in colour distribution, and intraclass similarity. Additionally, these techniques are limited to specific magnification levels of input images. As a result, the need to develop multiclass classification techniques independent of magnification levels has become increasingly demanding. This research introduces an innovative approach based on an enhanced vision transformer (ViT) for BC analysis that seamlessly handles binary and multiclass classification tasks in a magnification-independent manner. The proposed framework incorporates convolutional layers into the ViT model to simultaneously enhance its capability to capture local and global features. This innovative adjustment also empowers the model to capture complex patterns and subtle features by overcoming the issue of different magnification levels. Extensive experiments are conducted on the benchmark BreaKHis dataset to validate the effectiveness of the proposed framework. It achieves an accuracy of 89.43% for binary classification (benign vs. malignant) and 57.94% and 74.71% for multiclass benign and multiclass malignant subclass classification, respectively. The results confirm that the proposed framework surpasses the performance of the original ViT model and other state-of-the-art techniques. © The Author(s), under exclusive licence to Springer science+Business Media, LLC, part of Springer Nature 2024.

关键词： Diseases

来源：评论

学校读者我要写书评

暂无评论

Dependency Parsing Using Recurrent Neural Network on Myanmar Language 16th

Dependency Parsing Using Recurrent Neural Network on Myanmar...

引用

16th International Conference on Genetic and Evolutionary Computing, ICGEC 2024

作者： Win, Nwe Nwe Pa Pa, Win Department of Myanmar Nationalities’ Languages Ministry of Education NayPyitaw Myanmar Ministry of Science and Technology University of Computer Studied Yangon Myanmar

ISBN: (纸本)9789819615308

Word segmentation, part of speech of tagging and dependency parsing are the important role in Natural Language Processing (NLP). The POS and dependency parsing information are also necessary for NLP’s applications such as machine translation (MT), information retrieval (IR), etc. Although there are many research efforts in this process, there is still necessary to develop standard model for the Myanmar Language. This system uses Recurrent Neural Network (RNN) to keep away from errors and improve segmentation by utilizing POS data. In this paper, this system compares BILSTM and Hidden Markov Model (HMM) and the performances showed that precision, recall, F1 score, support and confusion matrix by using MLPOS. © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2025.

关键词： Hidden Markov models

来源：评论

学校读者我要写书评

暂无评论

Personalized Pronunciation Practice Recommendation in Non-native French Speakers Using Graph-Based Analysis 17th

Personalized Pronunciation Practice Recommendation in Non-na...

引用

17th International Conference on Multi-disciplinary Trends in Artificial Intelligence, MIWAI 2024

作者： Kuptavanich, Kittipitch Wita, Ratsameetip Phoungsub, Montiya Department of Computer Science Faculty of Science Chiang Mai University Chiang Mai Thailand Department of Western Languages Faculty of Humanities Chiang Mai University Chiang Mai Thailand

ISBN: (纸本)9789819606917

This paper presents a framework aimed at improving the pronunciation accuracy of non-native French speakers, particularly those preparing for the DELF A2 exam. It leverages advanced speech recognition and phonetic analysis to identify frequent pronunciation errors at both word and phoneme levels. A graph-based knowledge base captures common mispronunciation patterns, which guide personalized pronunciation lessons focusing on prevalent errors. The system has been introduced to ten second-year French program students from Chiang Mai University, and early results indicate improvements in pronunciation accuracy, especially in vowel distinction and reducing errors with similar-sounding words. Ongoing data collection will further refine the system to address the unique pronunciation challenges faced by Thai learners. This adaptive, data-driven approach has demonstrated potential for enhancing pronunciation fluency, with future work focusing on broader scalability and application to diverse learner populations. © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2025.

关键词： Contrastive Learning

来源：评论

学校读者我要写书评

暂无评论

IALS: Innovative Approach for Lung Segmentation Applying Artificial Intelligence and Deep Learning

IALS: Innovative Approach for Lung Segmentation Applying Art...

引用

2025 International Conference on computer, Electrical and Communication Engineering, ICCECE 2025

作者： Hossain, Mohammad Alamgir Shahi, Fazal Imam Husain, Sadia Suliman, Abdelrazig Kamal, Mohammad Shahid Ahmad, Yasir Talukdar, Mohammad Mazedul Huq Martin, R. John Jazan University College of Engineering & Computer Science Department of computer Science Jazan Saudi Arabia Jazan University ELearning Center Jazan Saudi Arabia University of Bisha Faculty of Computers and Information Technology Department of Information Technology Saudi Arabia Jazan University College of Arts and Humanities Department of Foreign Languages Jazan Saudi Arabia

ISBN: (纸本)9798331513894

The automated segmentation of lungs within chest CT scans remains crucial for many medical applications that require disease diagnosis and treatment design. The presented deep incremental learning (DIL) system addresses fundamental limitations that arise during lung segmentation processes. The exceptional segmentation outcomes result from the neural network structure which includes masked RCNN+UNet for tissue segmentation and GAN+VARMA for classification. Medical literature identifies lung tissue structures as difficult to study because they present irregular forms and complex structures. The task of obtaining precise segmentation proves very challenging. The application of masked RCNN+UNet enables shape control and improves segmentation precision of the lung system. The DIL function helps maintain improved features within segmented images. The applied framework enables improved performance through DSC and MAE and JI metrics evaluation. Experimental evidence shows that the new framework achieves superior performance compared to current top-ranking methods for lung segmentation. The proposed framework exhibits excellent precision and detection efficiency with a DSC value of 0.96 and JI value of 0.93 and a low MAE © 2025 IEEE.

关键词： Medical applications

来源：评论

学校读者我要写书评

暂无评论

Windows Registry Attack and Defense: Behavior Analysis and Persistence Detection Based on ETW 2nd

Windows Registry Attack and Defense: Behavior Analysis and P...

引用

2nd International Conference on Security and Information Technologies with AI, Internet Computing and Big-data Applications, SITAIBA 2023

作者： Chen, Jian-Xin Chang, Yang-Hsin Hwang, Yan-Ling Hunag, Jian-Hong Wang, Hao-Jyun Hsu, Fu-Hau Department of Computer Science and Information Engineering National Central University Taoyuan32001 Taiwan School of Applied Foreign Languages Chung Shan Medical University Taichung40201 Taiwan

ISBN: (纸本)9789819777853

Windows Registry contains various information about users and can be viewed as a database. While Microsoft gives users customization choices, it also unintentionally turns into a tool that attackers may fully use to carry out fileless assaults, incursion persistence, and other harmful actions. The opening section of this paper introduces the registry's logical structure and internal data structures. The following section examines registry-related threats and how the current Windows 10 operating system responds to them. Finally, it shows how a system based on Windows’ integrated logging system, ETW, has been implemented. The monitoring application records and filters registry write operation data, and the filtered data is then examined and sent to VirusTotal to ascertain if the written data was malicious. This study implements an ETW-based registry attack detection system that is effective and dependable. By integrating with VirusTotal, we can more quickly and correctly identify and stop registry assaults or misuse, protecting the security of user PCs. © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2025.

关键词： computer viruses

来源：评论

学校读者我要写书评

暂无评论

A Robust and Secure Framework for an AI-Assisted Academic Writing Platform 23rd

A Robust and Secure Framework for an AI-Assisted Academic...

引用

23rd International Conference on Web-Based Learning, ICWL 2024 was held in conjunction with the 9th International Symposium on Emerging Technologies for Education, SETE 2024

作者： Qureshi, Umair Mujtaba Tong, Esther Ka-Man Chen, Zili Val Deng, Chunrao Nandakumar, Gaya Division of Languages and Communication School of Professional Education and Executive Development The Hong Kong Polytechnic University China Computer Science and Engineering Department The Chinese University of Hong Kong China

ISBN: (纸本)9789819644063

The Artificial Intelligence (AI)-assisted academic writing platform is a digital tool that leverages the use of artificial intelligence (AI) technology to help the platform users, i.e., students, teachers, and professionals in various aspects of academic writing. This work reviews existing AI-assisted academic writing platforms, discusses core system design features that are important considerations in designing such frameworks, and provides a critical discussion of the strengths and weaknesses of existing platforms. This work also proposes a robust and secure framework for an AI-assisted academic writing platform based on Azure cloud services. The platform aims to provide comprehensive supervisory support, free from time and location constraints, to help its users e.g., teachers in terms of helping with their workload in assessing academic reports, and to students by providing support to enhance their academic writing. The proposed framework is based on Microsoft (MS) Azure cloud services because it provides reliability, scalability, and a wide range of services such as Web Server and Content Management System (CMS), Azure Database for MySQL and OpenAI Azure Service i.e., ChatGPT in a secure environment. In addition, the proposed framework integrates with a Private Cloud to further enhance the platform’s functionality and secure environment. The main objective of the platform is to deliver an interactive system that provides reliable feedback on English writing in a range of academic fields. The proposed platform is then evaluated by piloting on the PolyU SPEED academic course led by 16 teachers and tested on academic reports of 273 students. The survey-based evaluation results show that the proposed AI-assisted academic writing platform is easy to use and showcases the ability to identify issues related to specific genre-based content and coherence, provide assistance in grammar and vocabulary, and highlight issues in reference style with reasonably high accuracy.

关键词： Teaching

来源：评论

学校读者我要写书评

暂无评论

Neuro-Symbolic Classification of Basic Human Values 2nd

Neuro-Symbolic Classification of Basic Human Values

引用

2nd International Workshop on Value Engineering in Artificial Intelligence, VALE 2024

作者： De Giorgis, Stefano Lazzari, Nicolas Institute of Cognitive Sciences and Technologies National Research Council Rome Italy Department of Computer Science University of Pisa Pisa Italy Department of Modern Languages Literature and Cultures University of Bologna Bologna Italy

ISBN: (纸本)9783031854620

This work explores the integration of ontology-based reasoning and Machine Learning techniques for explainable classification in the domain of moral and cultural values. By relying on an ontological formalization of moral values as in the Basic Human Values theory, which is based on the Description and Situation Ontology Design Pattern, the sandra neuro-symbolic reasoner is used to infer values (formalized as descriptions) that are satisfied by a certain sentence. Sentences, alongside their structured representation, are automatically generated using an open-source Large Language Model. The inferred descriptions are used to automatically detect the value associated with a sentence. We show that only relying on the reasoner’s inference results in explainable classification comparable to other more complex approaches. Moreover, we test how the LLMs tacit knowledge can be exploited to obtain novel formalizations of the domain. We show that combining the reasoner’s inferences with simple distributional semantics methods largely outperforms all the baselines, including complex models based on neural network architectures. © The Author(s), under exclusive license to Springer Nature Switzerland AG 2025.

关键词： Knowledge graph

来源：评论

学校读者我要写书评

暂无评论

Scene and Aerial Image Classification Based on Advanced Deep Learning Technique 8th

Scene and Aerial Image Classification Based on Advanced Deep...

引用

8th World Conference on Smart Trends in Systems Security and Sustainability, WorldS4 2024

作者： Mehmood, Maryam Hussain, Farhan Ijaz, Asad Department of Computer and Software Engineering National University of Sciences and Technology Islamabad44000 Pakistan Department of Software Engineering National University of Modern Languages Islamabad Pakistan Department of Mechanical Engineering Mirpur University of Science and Technology MUST AJK Mirpur Pakistan

ISBN: (纸本)9789819786947

Classifying scenes and aerial imagery is a critical component in applications such as land-use analysis, land cover mapping, and remote sensing technologies. Numerous existing models leverage Convolutional Neural Networks for classification, but they often involve intricate designs with a vast number of parameters, necessitating extensive computational power and prolonged training periods. This paper introduces a refined strategy employing a pre-trained EfficientNet B0 framework. Modification of parameters and layers through careful fine-tuning has resulted in notable enhancements in the model's classification precision. The superiority of EfficientNet B0 over comparative CNN architectures was confirmed by testing with three disparate datasets, particularly within the sphere of remote sensing imagery. The EfficientNet B0's superior performance is credited to its optimal structure that strikes a balance between model complexity and efficiency. Effective fine-tuning coupled with judicious computational resource allocation yields substantial accuracy in categorization. Moreover, its adaptability to assimilate remote sensing data exemplifies its widespread applicability. This study accentuates the significance of refining prevailing deep learning models to cater to specified objectives. Tailoring pre-trained models such as EfficientNet B0 enables marked advancements in classification precision, circumventing the necessity for intricate systems or significant computational demands. This methodology is promising for elevating both the proficiency and impact of scene and aerial image classification across a range of fields, with particular implications for remote sensing endeavors. © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2025.

关键词： Resource allocation

来源：评论

学校读者我要写书评

暂无评论

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案：

请选择收藏分类：

通借通还

建议与咨询 留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案： 新增检索档案 确定 取消

请选择收藏分类： 新增自定义分类 确定 取消

通借通还

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

请选择保存的检索档案：

请选择收藏分类：