Article
Computer Science, Artificial Intelligence
Jinsong Su, Jialong Tang, Hui Jiang, Ziyao Lu, Yubin Ge, Linfeng Song, Deyi Xiong, Le Sun, Jiebo Luo
Summary: In this study, a progressive self-supervised attention learning approach is proposed to enhance the performance of aspect-based sentiment analysis (ABSA) models by continuously learning useful attention supervision information through iterative processes. Experimental results demonstrate that this method can effectively improve the attention mechanism of ABSA models and enhance their performance.
ARTIFICIAL INTELLIGENCE
(2021)
Article
Computer Science, Information Systems
Ruijie Zhao, Congyan Lang, Zun Li, Liqian Liang, Lili Wei, Songhe Feng, Tao Wang
Summary: Pedestrian attribute recognition is widely used in pedestrian tracking and re-identification. This study proposes a Cross Attribute and Feature Network (CAFN) to tackle the challenges of multi-label nature and data sample characteristics. The network exploits correlations between attributes and achieves superior performance compared to state-of-the-art approaches.
MULTIMEDIA SYSTEMS
(2022)
Article
Computer Science, Artificial Intelligence
Liang Peng, Yang Yang, Zheng Wang, Zi Huang, Heng Tao Shen
Summary: Visual Question Answering (VQA) is a task that aims to answer natural language questions about visual images. Existing approaches often use attention mechanisms to focus on relevant visual objects and consider the relationships between objects. However, these approaches have limitations in modeling complex object relationships and leveraging the cooperation between visual appearance and relationships. To address these issues, we propose a novel end-to-end VQA model, called Multi-modal Relation Attention Network (MRA-Net). The model combines textual and visual relations, utilizes self-guided word relation attention, and incorporates question-adaptive visual relation attention modules to improve performance and interpretability. Experimental results on multiple benchmark datasets demonstrate that our proposed model outperforms state-of-the-art approaches.
IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE
(2022)
Article
Chemistry, Analytical
Shangwu Hou, Gulanbaier Tuerhong, Mairidan Wushouer
Summary: In sentiment analysis, biased user reviews can negatively impact a company's evaluation. Identifying such users is highly beneficial as their reviews are not based on reality but on their psychological characteristics. Biased users may also contribute to the spread of prejudiced information on social media. Therefore, proposing a method to detect polarized opinions in product reviews offers significant advantages.
Article
Computer Science, Artificial Intelligence
Chen An, Xiaodong Wang, Zhiqiang Wei, Ke Zhang, Lei Huang
Summary: This paper presents a novel method called MSMGA-Net for fine-grained visual classification. It explores information at different granularities and scales to locate key discriminative regions and extract discriminative features. The experiments validate the effectiveness of MSMGA-Net, achieving high accuracy on benchmark datasets.
APPLIED SOFT COMPUTING
(2023)
Article
Chemistry, Multidisciplinary
Guangtao Xu, Peiyu Liu, Zhenfang Zhu, Jie Liu, Fuyong Xu
Summary: This paper introduces an aspect-based sentiment classification method based on attention-enhanced graph convolutional network (AEGCN), which combines semantic and syntactic information by introducing multi-head attention (MHA). Experimental results demonstrate that this method outperforms traditional methods in utilizing semantic and syntactic information.
APPLIED SCIENCES-BASEL
(2021)
Article
Chemistry, Analytical
Sergiu Cosmin Nistor, Mircea Moca, Darie Moldovan, Delia Beatrice Oprean, Razvan Liviu Nistor
Summary: This paper introduces a sentiment analysis solution on tweets using Recurrent Neural Networks, achieving an accuracy rate of 80.74% after experimenting with 20 design approaches. The solution integrates an attention mechanism and a two-way localization system, based on an in-depth literature review for Twitter sentiment analysis.
Article
Computer Science, Artificial Intelligence
Hao Zhang, Yanan Liu, Zhaoyu Xiong, Zhichao Wu, Dan Xu
Summary: This paper proposes a Transformer-based visual semantic correlation network for visual sentiment analysis. By using an extended attention network and an object query tool, it comprehensively considers the correlation among visual components and filters out redundant and noisy visual proposals. Experiments show that this method outperforms other methods on multiple datasets.
COMPLEX & INTELLIGENT SYSTEMS
(2023)
Article
Computer Science, Artificial Intelligence
Wenxiong Liao, Bi Zeng, Xiuwen Yin, Pengfei Wei
Summary: The proposed multi-task aspect-category sentiment analysis model based on RoBERTa outperforms other models by extracting features from text and aspect tokens effectively and enhancing performance through cross-attention mechanism in experiments.
APPLIED INTELLIGENCE
(2021)
Article
Computer Science, Artificial Intelligence
Xiaofei Zhu, Ling Zhu, Jiafeng Guo, Shangsong Liang, Stefan Dietze
Summary: The study introduces a Global and Local Dependency Guided Graph Convolutional Networks (GL-GCN) approach, which integrates global and local structure information into aspect-based sentiment classification. Experimental results demonstrate that this method outperforms existing approaches in terms of accuracy and F1-Score.
EXPERT SYSTEMS WITH APPLICATIONS
(2021)
Article
Computer Science, Hardware & Architecture
Wenxiong Liao, Bi Zeng, Jianqi Liu, Pengfei Wei, Xiaochun Cheng, Weiwen Zhang
Summary: This paper proposes a novel multi-level graph neural network (MLGNN) for text sentiment analysis, which combines local and global features and utilizes a scaled dot-product attention mechanism to fuse the features of each word node in the graph.
COMPUTERS & ELECTRICAL ENGINEERING
(2021)
Article
Computer Science, Information Systems
Wei Zhou, Zhiwu Xia, Peng Dou, Tao Su, Haifeng Hu
Summary: In this study, a Double Attention framework based on the Graph Attention Network (DA-GAT) is proposed to effectively learn the correlation between labels from training data. By designing a new channel attention mechanism and a new label attention mechanism, the semantic correlation between channel feature maps and the correlation between labels are enhanced, respectively. The fusion of the outputs of these two attention mechanisms further improves the model performance.
ACM TRANSACTIONS ON MULTIMEDIA COMPUTING COMMUNICATIONS AND APPLICATIONS
(2023)
Article
Computer Science, Hardware & Architecture
Xiaowen Li, Ran Lu, Peiyu Liu, Zhenfang Zhu
Summary: This paper proposes a method that utilizes a hierarchical multi-head attention mechanism and a graph convolutional network to address the problem of linking aspect words and context in sentiment classification. Extensive experiments validate the effectiveness of the proposed method.
JOURNAL OF SUPERCOMPUTING
(2022)
Article
Computer Science, Theory & Methods
Zhongnan Zhao, Wenjing Liu, Kun Wang
Summary: This paper proposes a mining method for public opinion sentiment analysis based on multi-model fusion transfer learning, which can improve the learning efficiency of sentiment features by integrating the advantages of different models. It introduces a transfer learning strategy and an attention mechanism to enhance the performance of the models.
JOURNAL OF BIG DATA
(2023)
Article
Computer Science, Artificial Intelligence
Wenbai Chen, Jingchen Li, Haobin Shi, Kao-Shing Hwang
Summary: This paper proposes an adaptive multi-sensor visual attention model (AM-MA) to enhance recurrent visual attention models. The AM-MA uses multiple sensors to observe the input and can adaptively add more sensors. It incorporates a self-evaluation mechanism and a fine-tune mechanism, eliminating the need for pre-training and achieving satisfactory results even with an inappropriate structure.
NEURAL COMPUTING & APPLICATIONS
(2022)
Article
Computer Science, Artificial Intelligence
Min Meng, Yiting Jacqueline Chua, Erwin Wouterson, Chin Peng Kelvin Ong
Article
Engineering, Electrical & Electronic
Min Meng, Xiaoyu Zhan
IEEE SIGNAL PROCESSING LETTERS
(2018)
Article
Computer Science, Artificial Intelligence
Min Meng, Jun Yu
IEEE TRANSACTIONS ON IMAGE PROCESSING
(2019)
Article
Computer Science, Artificial Intelligence
Min Meng, Yu Liu, Jigang Wu
Summary: A novel image classification method is proposed in this paper, which simultaneously considers margin and locality structure information based on low-rank and sparse representation for more robust and comprehensive classification. Extensive experiments on four databases demonstrate that the proposed method outperforms other state-of-the-art algorithms.
NEURAL PROCESSING LETTERS
(2021)
Article
Engineering, Electrical & Electronic
Min Meng, Mengcheng Lan, Jun Yu, Jigang Wu
Summary: This paper proposes a novel transfer learning approach that integrates domain invariant feature learning, discriminative structure preservation, and sample reweighting into a unified learning model. The method effectively addresses the negative transfer problem and has been experimentally verified to outperform other state-of-the-art algorithms.
IEEE TRANSACTIONS ON CIRCUITS AND SYSTEMS FOR VIDEO TECHNOLOGY
(2021)
Article
Engineering, Electrical & Electronic
Min Meng, Zhuanghui Wu, Tianyou Liang, Jun Yu, Jigang Wu
Summary: Unsupervised domain adaptation aims to address two challenging problems: thoroughly exploring fine-grained cluster structure knowledge in source and target domains, and effectively incorporating these structure knowledge for adaptation. The authors propose structural representation learning and a novel structural centroid-based label prediction method to tackle these problems, and adopt clustering learning to incorporate discriminative structure knowledge for adaptation. Comprehensive experiments demonstrate the superiority of the proposed framework.
IEEE TRANSACTIONS ON CIRCUITS AND SYSTEMS FOR VIDEO TECHNOLOGY
(2022)
Article
Computer Science, Information Systems
Min Meng, Mengcheng Lan, Jun Yu, Jigang Wu, Ligang Liu
Summary: Unsupervised domain adaptation is a technique that learns robust classifiers for unlabeled target domain by borrowing knowledge from a well-established source domain. However, previous works have limitations in terms of overfitting and conditional distribution matching. This paper proposes a Dual-Level Adaptive and Discriminative (DLAD) classifier learning framework that addresses these limitations and achieves better performance than other competitive methods.
IEEE TRANSACTIONS ON MULTIMEDIA
(2023)
Article
Computer Science, Artificial Intelligence
Zhuanghui Wu, Min Meng, Tianyou Liang, Jigang Wu
Summary: In this study, we propose a Hierarchical Triple-level Alignment (HTA) method for unsupervised multisource-multitarget domain adaptation (UMDA). We introduce a triple-level alignment mechanism to effectively reduce domain shift among multiple source and target domains. Our method can incorporate domain label, class label, and data structure information for effective knowledge transfer. Experimental results on three standard benchmarks demonstrate the superiority of the proposed framework.
APPLIED INTELLIGENCE
(2023)
Article
Automation & Control Systems
Min Meng, Mengcheng Lan, Jun Yu, Jigang Wu
Summary: This article introduces a novel multiview subspace learning method called multiview consensus structure discovery (MvCSD). Compared to existing methods, MvCSD is able to better utilize intrinsic complementary information to produce a more robust and accurate representation structure. The proposed method can be effectively optimized using an alternating iterative algorithm and eigendecomposition, with theoretical convergence guarantee.
IEEE TRANSACTIONS ON CYBERNETICS
(2022)
Article
Engineering, Electrical & Electronic
Mengcheng Lan, Min Meng, Jun Yu, Jigang Wu
Summary: In this paper, a novel Generalized Multi-view Collaborative Subspace Clustering (GMCSC) framework is introduced to address the problem of multi-view clustering. By jointly learning the consensus subspace structure of all views and the embedding subspaces for each view, and by using self-representation learning strategy and exploring complementary information, the proposed method effectively solves the problem of feature degeneration and multi-view data processing.
IEEE TRANSACTIONS ON CIRCUITS AND SYSTEMS FOR VIDEO TECHNOLOGY
(2022)
Article
Computer Science, Software Engineering
Jiaqi Guan, Min Meng, Tianyou Liang, Jigang Liu, Jigang Wu
Summary: Generalized zero-shot learning (GZSL) aims to recognize seen and unseen samples using semantic information. Existing approaches suffer from fine-grained cluster structure and overfitting issues. To address these challenges, we propose a Dual-level Contrastive Learning Network (DCLN) that integrates intra-domain and cross-domain contrastive learning.
Article
Computer Science, Artificial Intelligence
Min Meng, Jie Wei, Jigang Wu
Summary: This article proposes a neural network-based ZSL model that incorporates an attention mechanism to discover discriminative parts in images. By using a simple yet effective objective function and a multiple-attention scheme, the proposed approach achieves superior performance compared to existing methods on the CUB-2010-2011 dataset.
IEEE TRANSACTIONS ON COGNITIVE AND DEVELOPMENTAL SYSTEMS
(2022)
Article
Computer Science, Artificial Intelligence
Min Meng, Haitao Wang, Jun Yu, Hui Chen, Jigang Wu
Summary: This article introduces a novel supervised cross-modal hashing method ASCSH, which decomposes mapping matrices to exploit correlation between modalities and uses a discrete asymmetric framework to fully explore supervised information, solving binary constraint problems effectively.
IEEE TRANSACTIONS ON IMAGE PROCESSING
(2021)
Article
Computer Science, Artificial Intelligence
Min Meng, Mengcheng Lan, Jun Yu, Jigang Wu, Dapeng Tao
IEEE TRANSACTIONS ON IMAGE PROCESSING
(2020)