教学课程资源库 更多>>

关键词： similarity algorithms   natural language processing   computational linguistics  

摘要： In an era defined by vast volumes of digital text, the capacity to compare, interpret, and quantify textual similarity is a cornerstone of modern computational linguistics and natural language processing (NLP). Text similarity algorithms support critical applications in information retrieval, plagiarism detection, sentiment analysis, text summarization, and beyond. This paper provides a comprehensive survey and comparative analysis of established text similarity algorithms, including edit-distance-based metrics (Levenshtein and Damerau-Levenshtein), character-based measures (Jaro and Jaro-Winkler), local sequence alignment (Smith-Waterman), vector-based semantic measures (Cosine similarity), and methods reliant on subsequence statistics (N-gram similarity). Each algorithm is analyzed in terms of its underlying theoretical foundations, computational complexity, performance characteristics, and domain-specific suitability. While traditional approaches excel in correcting typographical errors or identifying subtle lexical variations, more robust methods handle semantically rich corpora, larger text bodies, and intricate linguistic phenomena. Moreover, potential avenues for improvement are explored, including hybridization of existing approaches and the integration of emerging machine learning and deep neural models. This holistic examination aims to inform the selection and development of text similarity measures for diverse real-world applications and to guide future research directions in computational linguistics.

关键词： Transformers   Feature extraction   Computational modeling   Superresolution   Image restoration   Image reconstruction   Computer science   Collaboration   Deep learning   Data mining  

摘要： Single image super-resolution (SR) can recover highresolution images from the corresponding low-resolution counterparts, which meets the application demands of consumer electronics, such as improving the visual experience of smart televisions (TVs) and virtual reality (VR) devices. Although deep learning-based SR methods have recently gained promising success, most existing approaches typically stack numerous network layers, which significantly increases the model complexity and hinders the deployment on electronics with limited computational capability. To tackle this problem, we propose a collaborative network of Mamba and CNN (CNMC) for lightweight image super-resolution. CNMC is mainly composed of multiple collaborative units of Mamba and CNN (CUMCs), which leverage the complementary advantages of Mamba and CNN to extract deep features beneficial for reconstruction. Specifically, CUMC introduces Mamba which enjoys the global receptive field and linear complexity, to perform long-range dependency modeling. Additionally, it employs CNN with the significant inductive bias to facilitate the local information interaction and compensation. The collaboration between Mamba and CNN effectively exploits both non-local and local priors, ensuring a comprehensive enhancement of deep features. Furthermore, a multi-scale spatial refinement attention (MSSRA) is developed in CUMC, to modulate channel-wise weights of features using a spatial fine-grained manner at different scales and then aggregate these cross-scale features, thereby distilling important information and restoring more precise details. Extensive quantitative and qualitative experiments demonstrate the superiority of our CNMC over other state-of-the-art lightweight SR methods. Most importantly, compared to the recent Transformer-based methods NGswin and HSSRNet, our CNMC achieves PSNR improvements of 0.21 dB and 0.32 dB for ×2 SR on Urban100 dataset. © 1975-2011 IEEE.

关键词： Emotion recognition   Computational modeling   Weak supervision   Multilingual   Benchmark testing   Text recognition   Speech recognition   Context modeling   Computer science   Complexity theory   deep learning   bidirectional encoder representations from transformers   weak supervision   neural machine translation   multi-view learning   German language  

摘要： Text-based emotion recognition plays a crucial role in various domains and applications due to its significance in understanding human behavior and improving communication systems. Although extensive research has been conducted for texts in English, only few studies deal with non-English languages. Especially for the German language, previous work is limited due to the focus on specific text domains and the usage of lexicon-based approaches, which lack the ability to consider contextual information. In this paper, an approach for emotion recognition on German text corpora is presented, which addresses these challenges with domain-generalized and context-sensitive methodologies. To achieve domain generalization, neural machine translation and weak supervision techniques are combined with multi-view learning, in which various data sources from different domains are utilized to improve the generalizability and to overcome the problem of lack of data. Using the BERT model as the primary architecture to capture contextual information, an overall F1-score of 65.5% and an accuracy of 68.1% are achieved while maintaining well-balanced results on the metrics over all considered domains and emotions. To ensure the legitimacy and reliability of the findings, further comprehensive evaluations and comparisons are conducted. Given the absence of an established benchmark specifically for emotion recognition on German text corpora, the used datasets, evaluations, and results can serve for benchmarking purposes.

关键词： Reinforcement learning   Correlation   Recommender systems   Trajectory   Standards   Computer science   Training   Streaming media   Q-learning   Multimedia systems   Label distribution learning   recommender systems   reinforcement learning  

摘要： Reinforcement learning (RL) aims to formulate the recommendation task as a Markov decision process (MDP) and trains an agent to automatically learn the optimal recommendation policy from interaction trajectories through trial-and-error and reward mechanisms. However, most existing RL-based approaches overlook the correlation between items and the dynamics of user interests implied in temporally close interactions. Therefore, in this paper, we propose a reinforcement learning method that incorporates a "recent-k items" distribution to capture users' local preferences. Specifically, we model the output layer as two distinct branches. The "recent-k items" branch, formulated with a Kullback-Leibler divergence loss, learns the recent interests of users, whereas the other branch utilizes a one-step temporal difference error to capture long-term preferences. The proposed structure is integrated into deep Q-learning and actor-critics, resulting in two enhanced methods named RkQ and RkAC, respectively. Furthermore, a novel soft inter-reward is carefully designed to enhance the proposed method, and we theoretically prove the convergence of the proposed algorithm. We perform extensive experiments on two large real-world datasets and conduct further analysis of the influences of different action sequences, time intervals, and enhancement capabilities for state-of-the-art models. The experimental results demonstrate the efficacy of our proposed methods.

关键词： Sentiment analysis  

摘要： In the information age, which is characterized by a proliferation of diverse data, big data technologies have emerged as a key driver for corporate digital transformation and improved enterprise performance. This study conducts an interdisciplinary investigation, integrating insights from computer science, linguistics, and management, to assess the importance of a positive tone in enhancing the impact of digital transformation on firm value. Advanced machine learning techniques and natural language processing algorithms are employed to analyze management discussion reports from listed Chinese companies (2007-2022). The research identifies keyword frequencies related to big data-driven digital transformation and analyses language sentiment using extensive lexicons and NLP models. Statistical models examine the reciprocal impact of digital transformation initiatives on enterprise value and the role of positive tone in reinforcing this relationship. Findings reveal significant improvements in enterprise value for enterprises engaging in big data-driven digital transformation, with positive tone amplifying these benefits. This research offers insights into synergies between technological adoption, communication strategies, and business performance. © 2025, Codon Publications. All rights reserved.

关键词： Decoding   Image segmentation   Medical diagnostic imaging   Feature extraction   Computer architecture   Semantics   Accuracy   Computer science   Training   Deep learning   Biomedical image processing   Medical image segmentation   medical imaging   deep learning  

摘要： Medical image segmentation entails assigning each pixel in an image to its corresponding class label, a challenging task given the considerable anatomical variations in different cases. The encoder-decoder approach, exemplified by architectures such as U-Net, has emerged as the predominant framework for medical imaging segmentation tasks. In recent years, diverse modifications to the U-Net architecture have been explored, giving rise to distinct models that showcase noteworthy results in comparison to the conventional U-Net design. In this paper, we introduce a novel architectural framework, which we refer to as the Polyhierarchical Residual Network (PolyRes-Net). Each encoder step comprises a Multi-Level Residual Block (MLR-block) designed to extract local and global feature maps. Furthermore, each decoder step is preceded by an attention gate, aiding in the extraction of the most salient features from the preceding layer, while skip connections correspond to the respective encoder steps. Lastly, the multi-scale feature aggregation (MSFA) block consolidates features from various decoder steps. Four benchamar datasets are used for evaluating our model: Krusir-SEG, CVC ClinicDB, 2018 Data Science Bowl, and ISIC-2018 skin lesion segmentation challenge dataset based on two metrics: the Mean Dice Similarity Coefficient (mDSC) and the Mean Intersection Over Union (mIOU). The results of the proposed PolyRes-Net outperformed the state-of-the-art segmentation methods. Specifically, PolyRes-Net achieves the highest mDSC scores of 91.02%, 91.80%, and 89.25% on CVC ClinicDB, 2018 Data Science Bowl, and ISIC-2018 skin lesion segmentation challenge dataset, respectively. Additionally, the highest mIOU scores are 85.60%, 85.32%, and 82.14% for the same datasets, further underscoring the efficacy of the proposed model.

关键词： Object recognition   Feature extraction   Reliability   Lesions   Information technology   Diseases   Data analysis   Computer science   Computer architecture   Accuracy  

关键词： Internet of Things   Training   Security   Biological system modeling   Data privacy   Data models   Botnet   Accuracy   Computer science   Servers   Consumer Internet of Things   federated learning   attack detection   marine predator algorithm   DDoS attacks  

摘要： The rising adoption of Consumer Internet of Things (CIoT) devices brings heightened concerns about security vulnerabilities and privacy breaches. Given their connection to daily routine and data collection abilities, it is vital to protect individual data against unauthorized entry and possible leaks proactively. Data protection's maximum communication and storage charge will gradually challenge the typical ecology of integrated over-the-cloud learning and handling for IoT environments. Federated learning (FL) is developed as the most reasonable method for this challenge. In FL, training data-driven machine learning (ML) approaches is a form of association among several users with no need for data carried to a fundamental point;therefore, improving communication and storage charge offers a great degree of consumer-level secrecy. FL is an advanced ML that performs appropriately by inherently prioritizing confidentiality, circumventing requirements for centralized data collection, and supporting the defence. FL improves security in IoT by enabling decentralized model training without centralizing sensitive data, preserving privacy, and reducing communication costs. Integrating FL into attack detection models ensures zero-trust security, minimizing vulnerabilities and enhancing protection against cyber threats while keeping data on-device. This study develops an enhanced artificial gorilla troop optimizer with an FL-based attack detection and classification (EGTO-FLADC) approach on the CIoT environment. The EGTO-FLADC approach aims to improve security in the CIoT environment via FL and the attack detection process. Primarily, the EGTO-FLADC technique utilizes the FL model for collaborative training of combined models with many devices with no data-sharing process. In addition, the EGTO-FLADC technique uses a min-max scalar for data normalization. For attack detection, the EGTO-FLADC technique employs the temporal convolution networks with a gated recurrent unit (TCN-GR

关键词： Fake news   Social networking (online)   Surveys   Reviews   Computer science   Feature extraction   Economics   Blogs   Voting   Source coding   Text-based fake news detection   image-based fake news detection   large language models   GPT4   systematic review   embeddings  

摘要： Social media platforms have transformed the world into a global village by providing a unique platform for unrestricted communication and opinion sharing. However, this freedom is used to spread misinformation and disrupt societal harmony. To combat misinformation and fake news on social media platforms, multifarious AI applications have been developed to detect such content in various languages and data modalities, including text, images, and videos. To establish a distinctive platform that fosters the rapid development of AI-driven fake news detectors, researchers have published several review articles in recent years. However, many of these articles are outdated, and lack comprehensiveness, particularly regarding recent trends, public datasets, representation learning methods, and classifiers details. The limited scope of articles hindered their ability to provide in-depth information on predictors' language specificity, data modality focus, and state-of-the-art performance across diverse languages. This paper offers a unique platform with detailed information on these aspects to address this gap. It offers a detailed roadmap for understanding the scope, strengths, and limitations of existing review articles on the subject. To support the development of new benchmark datasets and more accurate predictors, it conducts an in-depth analysis of the definitions and perspectives of fake news within the context of existing literature. This analysis provides a more comprehensive definition and basic concepts of fake news. In addition, 310 fake news detection articles published in the last 8 years have been thoroughly investigated. Within the landscape of these articles, the study presents details of languages, datasets, data modalities, predictor architecture designs, and their performance metrics.

关键词： Computer and information sciences   Schools   Human learning   Mathematics   Randomized controlled trials   Children   Robotics   Robots  

摘要： With the growing digitization of society, there is a need to enhance computational thinking as an indispensable skill for modern daily life. Consequently, computer science education for children at early ages has become increasingly important. This study conducts a randomized controlled trial to examine the impact of the interventions using educational robotics as well as computer-aided mathematics drills (via a "math app") on students' performance in primary schools in Basra, Iraq. We provide several new empirical findings. First, the short-run impact of robotics-based learning on computational thinking is positive and statistically significant for girls, particularly poor performing girls, but not for boys. Second, the impact on computational thinking is augmented by introducing a math app, further improving computational thinking. Together, these two interventions also enhance general intelligence. Third, the positive impact was still evident more than three months after the interventions for girls who received both computer science and math education, suggesting their complementarity. Our results show that computer science education using educational robots in primary schools is effective in enhancing computational thinking and relevant skills.