Article
Automation & Control Systems
Mengyao Lyu, Hu Han, Xiangzhi Bai
Summary: The goal of zero-shot learning is to transfer knowledge from seen to unseen classes by using auxiliary information. Most existing methods view ZSL as a label-embedding problem and face challenges such as bias towards seen classes and sacrificing performance. In this article, an embedding approach inspired by human recognition memory is proposed to effectively address these issues and outperform state-of-the-art methods.
IEEE TRANSACTIONS ON SYSTEMS MAN CYBERNETICS-SYSTEMS
(2022)
Article
Psychology, Multidisciplinary
David J. Hauser, Norbert Schwarz
Summary: Concepts often appear alongside related concepts in everyday language, and these collocations often reflect societal biases. This study examines the influence of collocation on implicit bias by testing the evaluation of neutral concepts that frequently collocate with valenced concepts. The results suggest that neutral concepts with valenced collocates have a similar effect on evaluation as strongly valenced concepts, indicating that collocations in natural language may contribute to the development of implicit bias.
FRONTIERS IN PSYCHOLOGY
(2022)
Article
Computer Science, Information Systems
Xiaofeng Ding, Tieyong Zeng, Jian Tang, Zhengping Che, Yaxin Peng
Summary: This paper proposes a novel semantic representation (SR) module for extracting semantic information in semantic segmentation tasks. The module enhances the representation ability of semantic context by utilizing global semantic information and improves the consistency of intraclass features by aggregating global features. Additionally, the SR module can be extended to build a semantic representation refinement network for enhancing the structural reasoning of the model through multiple-scale iterations.
IEEE TRANSACTIONS ON MULTIMEDIA
(2023)
Article
Computer Science, Software Engineering
P. K. Bhagat, Prakash Choudhary, Kh Manglem Singh
Summary: This article discusses the principles and methods of zero-shot learning, focusing on the different ways to construct semantic space and categorizing zero-shot learning methods accordingly. It also introduces the relevant performance evaluation measures and databases.
Article
Chemistry, Multidisciplinary
Huy Manh Nguyen, Tomo Miyazaki, Yoshihiro Sugaya, Shinichiro Omachi
Summary: A novel framework for visual-semantic embedding is proposed in this paper, which maps instances into multiple individual embedding spaces to capture multiple relationships between instances, leading to compelling video retrieval. The proposed method achieved superior performance in sentence-to-video retrieval experiments, demonstrating the effectiveness of the multiple embedding approach compared to existing methods.
APPLIED SCIENCES-BASEL
(2021)
Article
Computer Science, Artificial Intelligence
Chenwei Tang, Zhenan He, Yunxia Li, Jiancheng Lv
Summary: This article proposes a structure-aligned generative adversarial network framework to address issues in zero-shot learning, such as semantic gap, domain shift, and hubness problem. The framework consists of a generative adversarial network with a softmax classifier part and a structure-aligned module to bridge the semantic gap between visual and semantic spaces. By generating pseudovisual features and aligning structures, classification performance is improved.
IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS
(2022)
Article
Engineering, Electrical & Electronic
Hongguang Zhu, Chunjie Zhang, Yunchao Wei, Shujuan Huang, Yao Zhao
Summary: Effective and efficient image-text retrieval is a challenging problem due to the gap between vision and language modalities. This study proposes a trade-off approach that balances effectiveness and efficiency by introducing the ESA module and SEL method. Experimental results demonstrate the effectiveness of the proposed method and its advantages in performance and retrieval time compared to other methods.
IEEE TRANSACTIONS ON CIRCUITS AND SYSTEMS FOR VIDEO TECHNOLOGY
(2023)
Article
Computer Science, Artificial Intelligence
Ning An, Meng Chen, Li Lian, Peng Li, Kai Zhang, Xiaohui Yu, Yilong Yin
Summary: This study focuses on the interpretability of venue representations and proposes two novel models, CEM and XEM, which can generate easy-to-understand venue representations. Experimental results demonstrate that the interpretability introduced to the venue representations improves the performance of various downstream tasks.
KNOWLEDGE-BASED SYSTEMS
(2022)
Article
Multidisciplinary Sciences
Shengzi Sun, Binghui Guo, Zhilong Mi, Zhiming Zheng
Summary: The study proposes a cross-modal semantic autoencoder, which retains semantic information by mapping data to a shared space, achieving significant progress in multi-modal retrieval.
SCIENTIFIC REPORTS
(2021)
Article
Computer Science, Information Systems
Yan Wang, Honghao Liang, Kaihong Zheng, Jingfeng Yang, Lukun Zeng, Qihang Gong, Sheng Li, Shangli Zhou
Summary: This paper proposes a knowledge graph embedding learning framework combined with triple semantic information (KGSE). KGSE comprehensively considers the structural embedding and semantic embedding of triples, and employs improved TransD model and deep convolutional neural network model to obtain the embedding representation. Experimental results show that the proposed framework improves significantly compared with Trans-based models and other baseline models in link prediction and triple classification tasks, verifying the effectiveness of the proposed framework.
Article
Mathematics
Wael Etaiwi, Arafat Awajan
Summary: This study proposes a novel semantic graph embedding-based abstractive text summarization technique for the Arabic language, namely SemG-TS, and demonstrates its superiority through experiments.
Article
Computer Science, Hardware & Architecture
Yue Ding, Yu-He Guo, Wei Lu, Hai-Xiang Li, Mei-Hui Zhang, Hui Li, An-Qun Pan, Xiao-Yong Du
Summary: Identifying attribute semantic types (AST) in relationships is crucial for data cleaning, schema matching, and database keyword search. This paper proposes a context-aware method to determine ASTs by transforming the problem into a multi-class classification and introducing a schema context aware model to predict ASTs.
JOURNAL OF COMPUTER SCIENCE AND TECHNOLOGY
(2023)
Article
Computer Science, Artificial Intelligence
Zhigang Hao, Wolfgang Mayer, Jingbo Xia, Guoliang Li, Li Qin, Zaiwen Feng
Summary: This paper proposes a new method based on ontology embedding incorporating the semantic and structural features for ontology alignment. The method is used to align two widely used food ontologies and three Chinese food classification ontologies. Experimental results show that this method enhances the performance compared to other advanced alignment systems, demonstrating the importance of learning semantic representation and structural representation.
JOURNAL OF WEB SEMANTICS
(2023)
Article
Computer Science, Information Systems
Behzad Naderalvojoud, Adnan Ozsoy
Summary: This paper proposes a non-sequential refinement approach for word embedding models, using a string matching algorithm to improve the vectors of particular words. By changing the training order and incorporating GPU-based string matching, the accuracy of word vectors and the training process can be enhanced.
CLUSTER COMPUTING-THE JOURNAL OF NETWORKS SOFTWARE TOOLS AND APPLICATIONS
(2021)
Article
Computer Science, Software Engineering
Wen Lv, Hongbo Shi, Shuai Tan, Bing Song, Yang Tao
Summary: Zero-shot object detection (ZSD) relies on semantic information to identify and localize novel classes. However, the fixed semantic embedding used in ZSD algorithms leads to a significant gap between visual and semantic spaces. To bridge this gap, a dynamic semantic knowledge graph (DSKG) is proposed, which utilizes a semantic knowledge graph to establish connections between categories and introduces a dynamic semantic reasoning mechanism to update semantic embedding. Experimental results demonstrate that DSKG achieves significant improvements on MS-COCO and PASCAL VOC datasets.