Jun Li
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Jun Li has a total of 42 co-authors. Here are the top co-authors by number of shared publications:
Haibo Wu
3 shared publications- Xiangyu Zhao
City University of Hong Kong, Kowloon Tong, Hong Kong
3 shared publications - Xinyan Su
Chinese Academy of Sciences
3 shared publications - Zhiheng Zhang
Tsinghua University
2 shared publications Chao Tong
School of Computer Science and Engineering, Beihang University, Beijing, China
1 shared publicationDeren Chen
College of Computer Science, Zhejiang University,HangzhouChina
1 shared publication
Latest Publications
- 22 June 2025
- Published by Springer Nature
- p. 73-86
Dynamic Group Effects Analysis for Online Influence Maximization in Hypergraphs
- 22 June 2025
- Published by Springer Nature
- p. 73-86
Dynamic Group Effects Analysis for Online Influence Maximization in Hypergraphs
- Open Access
- 1 January 2025
- Published by Tech Science Press in Computers, Materials & Continua
- Vol. 82 (2) , 3381-3397
External Knowledge-Enhanced Cross-Attention Fusion Model for Tobacco Sentiment Analysis
In the age of information explosion and artificial intelligence, sentiment analysis tailored for the tobacco industry has emerged as a pivotal avenue for cigarette manufacturers to enhance their tobacco products. Existing solutions have primarily focused on intrinsic features within consumer reviews and achieved significant progress through deep feature extraction models. However, they still face these two key limitations: (1) neglecting the influence of fundamental tobacco information on analyzing the sentiment inclination of consumer reviews, resulting in a lack of consistent sentiment assessment criteria across thousands of tobacco brands; (2) overlooking the syntactic dependencies between Chinese word phrases and the underlying impact of sentiment scores between word phrases on sentiment inclination determination. To tackle these challenges, we propose the External Knowledge-enhanced Cross-Attention Fusion model, CITSA. Specifically, in the Cross Infusion Layer, we fuse consumer comment information and tobacco fundamental information through interactive attention mechanisms. In the Textual Attention Enhancement Layer, we introduce an emotion-oriented syntactic dependency graph and incorporate sentiment-syntactic relationships into consumer comments through a graph convolution network module. Subsequently, the Textual Attention Layer is introduced to combine these two feature representations. Additionally, we compile a Chinese-oriented tobacco sentiment analysis dataset, comprising 55,096 consumer reviews and 2074 tobacco fundamental information entries. Experimental results on our self-constructed datasets consistently demonstrate that our proposed model outperforms state-of-the-art methods in terms of accuracy, precision, recall, and F1-score.
- 1 January 2025
- Published by Tech Science Press in Computers, Materials & Continua
- Vol. 82 (2) , 3381-3397
Open AccessExternal Knowledge-Enhanced Cross-Attention Fusion Model for Tobacco Sentiment Analysis
In the age of information explosion and artificial intelligence, sentiment analysis tailored for the tobacco industry has emerged as a pivotal avenue for cigarette manufacturers to enhance their tobacco products. Existing solutions have primarily focused on intrinsic features within consumer reviews and achieved significant progress through deep feature extraction models. However, they still face these two key limitations: (1) neglecting the influence of fundamental tobacco information on analyzing the sentiment inclination of consumer reviews, resulting in a lack of consistent sentiment assessment criteria across thousands of tobacco brands; (2) overlooking the syntactic dependencies between Chinese word phrases and the underlying impact of sentiment scores between word phrases on sentiment inclination determination. To tackle these challenges, we propose the External Knowledge-enhanced Cross-Attention Fusion model, CITSA. Specifically, in the Cross Infusion Layer, we fuse consumer comment information and tobacco fundamental information through interactive attention mechanisms. In the Textual Attention Enhancement Layer, we introduce an emotion-oriented syntactic dependency graph and incorporate sentiment-syntactic relationships into consumer comments through a graph convolution network module. Subsequently, the Textual Attention Layer is introduced to combine these two feature representations. Additionally, we compile a Chinese-oriented tobacco sentiment analysis dataset, comprising 55,096 consumer reviews and 2074 tobacco fundamental information entries. Experimental results on our self-constructed datasets consistently demonstrate that our proposed model outperforms state-of-the-art methods in terms of accuracy, precision, recall, and F1-score.
- 30 November 2023
- Published by Springer Nature
- p. 295-308
Causal-Inspired Influence Maximization in Hypergraphs Under Temporal Constraints
- 30 November 2023
- Published by Springer Nature
- p. 295-308
Causal-Inspired Influence Maximization in Hypergraphs Under Temporal Constraints
- Open Access
Adversarial Batch Inverse Reinforcement Learning: Learn to Reward from Imperfect Demonstration for Interactive Recommendation
Rewards serve as a measure of user satisfaction and act as a limiting factor in interactive recommender systems. In this research, we focus on the problem of learning to reward (LTR), which is fundamental to reinforcement learning. Previous approaches either introduce additional procedures for learning to reward, thereby increasing the complexity of optimization, or assume that user-agent interactions provide perfect demonstrations, which is not feasible in practice. Ideally, we aim to employ a unified approach that optimizes both the reward and policy using compositional demonstrations. However, this requirement presents a challenge since rewards inherently quantify user feedback on-policy, while recommender agents approximate off-policy future cumulative valuation. To tackle this challenge, we propose a novel batch inverse reinforcement learning paradigm that achieves the desired properties. Our method utilizes discounted stationary distribution correction to combine LTR and recommender agent evaluation. To fulfill the compositional requirement, we incorporate the concept of pessimism through conservation. Specifically, we modify the vanilla correction using Bellman transformation and enforce KL regularization to constrain consecutive policy updates. We use two real-world datasets which represent two compositional coverage to conduct empirical studies, the results also show that the proposed method relatively improves both effectiveness (2.3\%) and efficiency (11.53\%)
Open AccessAdversarial Batch Inverse Reinforcement Learning: Learn to Reward from Imperfect Demonstration for Interactive Recommendation
Rewards serve as a measure of user satisfaction and act as a limiting factor in interactive recommender systems. In this research, we focus on the problem of learning to reward (LTR), which is fundamental to reinforcement learning. Previous approaches either introduce additional procedures for learning to reward, thereby increasing the complexity of optimization, or assume that user-agent interactions provide perfect demonstrations, which is not feasible in practice. Ideally, we aim to employ a unified approach that optimizes both the reward and policy using compositional demonstrations. However, this requirement presents a challenge since rewards inherently quantify user feedback on-policy, while recommender agents approximate off-policy future cumulative valuation. To tackle this challenge, we propose a novel batch inverse reinforcement learning paradigm that achieves the desired properties. Our method utilizes discounted stationary distribution correction to combine LTR and recommender agent evaluation. To fulfill the compositional requirement, we incorporate the concept of pessimism through conservation. Specifically, we modify the vanilla correction using Bellman transformation and enforce KL regularization to constrain consecutive policy updates. We use two real-world datasets which represent two compositional coverage to conduct empirical studies, the results also show that the proposed method relatively improves both effectiveness (2.3\%) and efficiency (11.53\%)
- Open Access
A General Neural Causal Model for Interactive Recommendation
Survivor bias in observational data leads the optimization of recommender systems towards local optima. Currently most solutions re-mines existing human-system collaboration patterns to maximize longer-term satisfaction by reinforcement learning. However, from the causal perspective, mitigating survivor effects requires answering a counterfactual problem, which is generally unidentifiable and inestimable. In this work, we propose a neural causal model to achieve counterfactual inference. Specifically, we first build a learnable structural causal model based on its available graphical representations which qualitatively characterizes the preference transitions. Mitigation of the survivor bias is achieved though counterfactual consistency. To identify the consistency, we use the Gumbel-max function as structural constrains. To estimate the consistency, we apply reinforcement optimizations, and use Gumbel-Softmax as a trade-off to get a differentiable function. Both theoretical and empirical studies demonstrate the effectiveness of our solution.
Open AccessA General Neural Causal Model for Interactive Recommendation
Survivor bias in observational data leads the optimization of recommender systems towards local optima. Currently most solutions re-mines existing human-system collaboration patterns to maximize longer-term satisfaction by reinforcement learning. However, from the causal perspective, mitigating survivor effects requires answering a counterfactual problem, which is generally unidentifiable and inestimable. In this work, we propose a neural causal model to achieve counterfactual inference. Specifically, we first build a learnable structural causal model based on its available graphical representations which qualitatively characterizes the preference transitions. Mitigation of the survivor bias is achieved though counterfactual consistency. To identify the consistency, we use the Gumbel-max function as structural constrains. To estimate the consistency, we apply reinforcement optimizations, and use Gumbel-Softmax as a trade-off to get a differentiable function. Both theoretical and empirical studies demonstrate the effectiveness of our solution.
- Open Access
- 21 October 2023
- Published by Association for Computing Machinery (ACM)
- p. 4115-4119
Counterfactual Adversarial Learning for Recommendation
Long-term user responses, i.e., clicks or purchases on e-commerce platforms, are crucial for sequential recommender systems. Recent off-policy evaluation methods involve these responses by simultaneously maximizing expected cumulative rewards. However, two aspects of these methods require further consideration. Firstly, from the system's point of view, candidates with various values are interchangeable, which may result in contradictory future recommendations despite having the same interaction history. Secondly, rewards are manually designed, which necessitates a trial-and-error approach to strike a balance between training stabilization and reward distinction. To address these issues, we propose a new sequential recommender system called NCM4Rec. Specifically, for the distinction problem, NCM4Rec achieves counterfactual consistency via a neural causal model, which is learnable yet equally expressive as classic structural causal models. Such consistency is maintained by a Gumbel-Max design. For the representing problem, NCM4Rec encodes different types of responses as one-hot vectors and captures the long-term preference via adversarial learning. As a consequence, NCM4Rec is both adaptive and identifiable. Both theoretical analyses of the consistency and empirical studies over two real-world datasets demonstrate the effectiveness of our method.
- 21 October 2023
- Published by Association for Computing Machinery (ACM)
- p. 4115-4119
Open AccessCounterfactual Adversarial Learning for Recommendation
Long-term user responses, i.e., clicks or purchases on e-commerce platforms, are crucial for sequential recommender systems. Recent off-policy evaluation methods involve these responses by simultaneously maximizing expected cumulative rewards. However, two aspects of these methods require further consideration. Firstly, from the system's point of view, candidates with various values are interchangeable, which may result in contradictory future recommendations despite having the same interaction history. Secondly, rewards are manually designed, which necessitates a trial-and-error approach to strike a balance between training stabilization and reward distinction. To address these issues, we propose a new sequential recommender system called NCM4Rec. Specifically, for the distinction problem, NCM4Rec achieves counterfactual consistency via a neural causal model, which is learnable yet equally expressive as classic structural causal models. Such consistency is maintained by a Gumbel-Max design. For the representing problem, NCM4Rec encodes different types of responses as one-hot vectors and captures the long-term preference via adversarial learning. As a consequence, NCM4Rec is both adaptive and identifiable. Both theoretical analyses of the consistency and empirical studies over two real-world datasets demonstrate the effectiveness of our method.
- 28 March 2023
- Published by Elsevier in Computer Communications
- Vol. 204, 147-157
PTF: Popularity-Topology-Freshness-based caching strategy for ICN-IoT networks
- 28 March 2023
- Published by Elsevier in Computer Communications
- Vol. 204, 147-157
PTF: Popularity-Topology-Freshness-based caching strategy for ICN-IoT networks
- 24 January 2023
- Published by Springer Nature in Peer-to-Peer Networking and Applications
- Vol. 16 (2) , 858-875
FabricETP: A high-throughput blockchain optimization solution for resolving concurrent conflicting transactions
- 24 January 2023
- Published by Springer Nature in Peer-to-Peer Networking and Applications
- Vol. 16 (2) , 858-875
FabricETP: A high-throughput blockchain optimization solution for resolving concurrent conflicting transactions
- 7 September 2022
- Published by Springer Nature
- p. 282-293
Multi-task Alignment Scheme for Span-level Aspect Sentiment Triplet Extraction
- 7 September 2022
- Published by Springer Nature
- p. 282-293
Multi-task Alignment Scheme for Span-level Aspect Sentiment Triplet Extraction
- Open Access
- 5 June 2018
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 6, 32754-32768
Design and Evaluation of Probabilistic Caching in Information-Centric Networking
In-network caching is a key feature of Information-Centric Networking (ICN). However, challenges still exist in ICN caching such as how to place content replicas among cache nodes to maximize cache system benefits without introducing too much overhead. In this paper, we formulate the content placement problem and propose a distributed probabilistic caching strategy to enhance cache efficiency. Each node makes cache decision individually and caches passing content with certain probability, which is proportional to content popularity and content placement benefit. As a component of our scheme, we also propose an accurate method to predict the variations of content popularity. Besides, a global-popularity-based caching scheme is proposed to be used as a benchmark for performance evaluation. We conduct extensive simulations based on ndnSIM and evaluate our scheme on tree, intra-AS and inter-AS topologies. Results indicate it outperforms the state-of-art schemes in terms of cache hit ratio, access latency, cache operation cost and link bandwidth savings. It can achieve dramatic performance improvement, even in the case of a small cache size. In particular, the reduction of caching operation can reach up to two orders of magnitude. We also examine the impacts of various replacement policies on cache performance and perform overhead analysis. Finally, we give a simplified implementation of our scheme and validate it via simulations.
- 5 June 2018
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 6, 32754-32768
Open AccessDesign and Evaluation of Probabilistic Caching in Information-Centric Networking
In-network caching is a key feature of Information-Centric Networking (ICN). However, challenges still exist in ICN caching such as how to place content replicas among cache nodes to maximize cache system benefits without introducing too much overhead. In this paper, we formulate the content placement problem and propose a distributed probabilistic caching strategy to enhance cache efficiency. Each node makes cache decision individually and caches passing content with certain probability, which is proportional to content popularity and content placement benefit. As a component of our scheme, we also propose an accurate method to predict the variations of content popularity. Besides, a global-popularity-based caching scheme is proposed to be used as a benchmark for performance evaluation. We conduct extensive simulations based on ndnSIM and evaluate our scheme on tree, intra-AS and inter-AS topologies. Results indicate it outperforms the state-of-art schemes in terms of cache hit ratio, access latency, cache operation cost and link bandwidth savings. It can achieve dramatic performance improvement, even in the case of a small cache size. In particular, the reduction of caching operation can reach up to two orders of magnitude. We also examine the impacts of various replacement policies on cache performance and perform overhead analysis. Finally, we give a simplified implementation of our scheme and validate it via simulations.