Kun Xiang
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Kun Xiang has a total of 4 co-authors. Here are the top co-authors by number of shared publications:
- Akihiro Fujii
Hosei University
2 shared publications Akihiro Fujii
Hosei University
1 shared publication- Akihiro Fujii
Applied Informatics, Hosei University, Tokyo, Japan
1 shared publication - Haochen Zou
School of Computer Science and Engineering, Nanjing University of Science and Technology, Nanjing, China
1 shared publication
Latest Publications
- 4 June 2026
- Published by World Scientific Pub Co Pte Ltd in International Journal of Semantic Computing
- Vol. 20 (02) , 227-248
Progressive Learning-to-Learn Knowledge Distillation for Robust Out-of-Domain Generalization
Knowledge distillation (KD) plays a crucial role in reducing computational costs, accelerating inference, and improving model generalization. However, when there is a significant capacity gap between teacher and student networks, KD often struggles to transfer knowledge effectively, and robustness under distribution shifts remains a major challenge. To address these issues, we propose learning-to-learn knowledge distillation (L2L-KD), a dynamic temperature-controlled KD framework that progressively increases learning difficulty, mimicking how human learners advance from basic to complex concepts. To further enhance robustness and generalization, we introduce a counterfactual data augmentation technique that leverages the Metropolis–Hastings algorithm to generate fluent and semantically coherent out-of-domain (OOD) samples. We evaluate L2L-KD across in-domain, OOD, and adversarial scenarios, and the results show that it consistently outperforms existing KD approaches while substantially improving robustness. Moreover, building upon this foundation, we extend the core learning philosophy to a new unsupervised cross-domain framework, demonstrating that the dynamic distillation principles of L2L-KD can naturally generalize to broader domain adaptation tasks.
- 4 June 2026
- Published by World Scientific Pub Co Pte Ltd in International Journal of Semantic Computing
- Vol. 20 (02) , 227-248
Progressive Learning-to-Learn Knowledge Distillation for Robust Out-of-Domain Generalization
Knowledge distillation (KD) plays a crucial role in reducing computational costs, accelerating inference, and improving model generalization. However, when there is a significant capacity gap between teacher and student networks, KD often struggles to transfer knowledge effectively, and robustness under distribution shifts remains a major challenge. To address these issues, we propose learning-to-learn knowledge distillation (L2L-KD), a dynamic temperature-controlled KD framework that progressively increases learning difficulty, mimicking how human learners advance from basic to complex concepts. To further enhance robustness and generalization, we introduce a counterfactual data augmentation technique that leverages the Metropolis–Hastings algorithm to generate fluent and semantically coherent out-of-domain (OOD) samples. We evaluate L2L-KD across in-domain, OOD, and adversarial scenarios, and the results show that it consistently outperforms existing KD approaches while substantially improving robustness. Moreover, building upon this foundation, we extend the core learning philosophy to a new unsupervised cross-domain framework, demonstrating that the dynamic distillation principles of L2L-KD can naturally generalize to broader domain adaptation tasks.
- Open Access
- 17 October 2023
- Published by MDPI AG in Big Data and Cognitive Computing
- Vol. 7 (4) , 166
HAMCap: A Weak-Supervised Hybrid Attention-Based Capsule Neural Network for Fine-Grained Climate Change Debate Analysis
Climate change (CC) has become a central global topic within the multiple branches of social disciplines. Natural Language Processing (NLP) plays a superior role since it has achieved marvelous accomplishments in various application scenarios. However, CC debates are ambiguous and complicated to interpret even for humans, especially when it comes to the aspect-oriented fine-grained level. Furthermore, the lack of large-scale effective labeled datasets is always a plight encountered in NLP. In this work, we propose a novel weak-supervised Hybrid Attention Masking Capsule Neural Network (HAMCap) for fine-grained CC debate analysis. Specifically, we use vectors with allocated different weights instead of scalars, and a hybrid attention mechanism is designed in order to better capture and represent information. By randomly masking with a Partial Context Mask (PCM) mechanism, we can better construct the internal relationship between the aspects and entities and easily obtain a large-scale generated dataset. Considering the uniqueness of linguistics, we propose a Reinforcement Learning-based Generator-Selector mechanism to automatically update and select data that are beneficial to model training. Empirical results indicate that our proposed ensemble model outperforms baselines on downstream tasks with a maximum of 50.08% on accuracy and 49.48% on F1 scores. Finally, we draw interpretable conclusions about the climate change debate, which is a widespread global concern.
- 17 October 2023
- Published by MDPI AG in Big Data and Cognitive Computing
- Vol. 7 (4) , 166
Open AccessHAMCap: A Weak-Supervised Hybrid Attention-Based Capsule Neural Network for Fine-Grained Climate Change Debate Analysis
Climate change (CC) has become a central global topic within the multiple branches of social disciplines. Natural Language Processing (NLP) plays a superior role since it has achieved marvelous accomplishments in various application scenarios. However, CC debates are ambiguous and complicated to interpret even for humans, especially when it comes to the aspect-oriented fine-grained level. Furthermore, the lack of large-scale effective labeled datasets is always a plight encountered in NLP. In this work, we propose a novel weak-supervised Hybrid Attention Masking Capsule Neural Network (HAMCap) for fine-grained CC debate analysis. Specifically, we use vectors with allocated different weights instead of scalars, and a hybrid attention mechanism is designed in order to better capture and represent information. By randomly masking with a Partial Context Mask (PCM) mechanism, we can better construct the internal relationship between the aspects and entities and easily obtain a large-scale generated dataset. Considering the uniqueness of linguistics, we propose a Reinforcement Learning-based Generator-Selector mechanism to automatically update and select data that are beneficial to model training. Empirical results indicate that our proposed ensemble model outperforms baselines on downstream tasks with a maximum of 50.08% on accuracy and 49.48% on F1 scores. Finally, we draw interpretable conclusions about the climate change debate, which is a widespread global concern.
- Open Access
DARE: Distill and Reinforce Ensemble Neural Networks for Climate-Domain Processing
Natural-language processing is well positioned to help stakeholders study the dynamics of ambiguous Climate Change-related (CC) information. Recently, deep neural networks have achieved good results on a variety of NLP tasks depending on high-quality training data and complex and exquisite frameworks. This raises two dilemmas: (1) the networks are highly reliant on powerful hardware devices and processing is time-consuming, which is not only inconducive to execution on edge devices but also leads to resource consumption. (2) Obtaining large-scale effective annotated data is difficult and laborious, especially when it comes to a special domain such as CC. In this paper, we propose a CC-domain-adapted BERT distillation and reinforcement ensemble (DARE) model for tackling the problems above. Specifically, we propose a novel data-augmentation strategy which is a Generator-Reinforced Selector collaboration network for countering the dilemma of CC-related data scarcity. Extensive experimental results demonstrate that our proposed method outperforms baselines with a maximum of 26.83% on SoTA and 50.65× inference time speed-up. Furthermore, as a remedy for the lack of CC-related analysis in the NLP community, we also provide some interpretable conclusions for this global concern.
Open AccessDARE: Distill and Reinforce Ensemble Neural Networks for Climate-Domain Processing
Natural-language processing is well positioned to help stakeholders study the dynamics of ambiguous Climate Change-related (CC) information. Recently, deep neural networks have achieved good results on a variety of NLP tasks depending on high-quality training data and complex and exquisite frameworks. This raises two dilemmas: (1) the networks are highly reliant on powerful hardware devices and processing is time-consuming, which is not only inconducive to execution on edge devices but also leads to resource consumption. (2) Obtaining large-scale effective annotated data is difficult and laborious, especially when it comes to a special domain such as CC. In this paper, we propose a CC-domain-adapted BERT distillation and reinforcement ensemble (DARE) model for tackling the problems above. Specifically, we propose a novel data-augmentation strategy which is a Generator-Reinforced Selector collaboration network for countering the dilemma of CC-related data scarcity. Extensive experimental results demonstrate that our proposed method outperforms baselines with a maximum of 26.83% on SoTA and 50.65× inference time speed-up. Furthermore, as a remedy for the lack of CC-related analysis in the NLP community, we also provide some interpretable conclusions for this global concern.
- Open Access
Sentiment Classification Method Based on Blending of Emoticons and Short Texts
With the development of Internet technology, short texts have gradually become the main medium for people to obtain information and communicate. Short text reduces the threshold of information production and reading by virtue of its short length, which is in line with the trend of fragmented reading in the context of the current fast-paced life. In addition, short texts contain emojis to make the communication immersive. However, short-text content means it contains relatively little information, which is not conducive to the analysis of sentiment characteristics. Therefore, this paper proposes a sentiment classification method based on the blending of emoticons and short-text content. Emoticons and short-text content are transformed into vectors, and the corresponding word vector and emoticon vector are connected into a sentencing matrix in turn. The sentence matrix is input into a convolution neural network classification model for classification. The results indicate that, compared with existing methods, the proposed method improves the accuracy of analysis.
Open AccessSentiment Classification Method Based on Blending of Emoticons and Short Texts
With the development of Internet technology, short texts have gradually become the main medium for people to obtain information and communicate. Short text reduces the threshold of information production and reading by virtue of its short length, which is in line with the trend of fragmented reading in the context of the current fast-paced life. In addition, short texts contain emojis to make the communication immersive. However, short-text content means it contains relatively little information, which is not conducive to the analysis of sentiment characteristics. Therefore, this paper proposes a sentiment classification method based on the blending of emoticons and short-text content. Emoticons and short-text content are transformed into vectors, and the corresponding word vector and emoticon vector are connected into a sentencing matrix in turn. The sentence matrix is input into a convolution neural network classification model for classification. The results indicate that, compared with existing methods, the proposed method improves the accuracy of analysis.
- 4 August 2021
- Published by Springer Nature
- p. 51-71
Reputation Analysis Based on Weakly-Supervised Bi-LSTM-Attention Network
- 4 August 2021
- Published by Springer Nature
- p. 51-71
Reputation Analysis Based on Weakly-Supervised Bi-LSTM-Attention Network