Hanqing Zhang
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
Research Subjects
Hanqing Zhang has a total of 9 co-authors. Here are the top co-authors by number of shared publications:
- Dawei Song
Beijing Institute of Technology, Beijing, China
2 shared publications Dawei Song
School of computer science & technology, Beijing Institute of Technology, Beijing, China
1 shared publication- Haolin Song
Beijing Institute of Technology, China
1 shared publication Limin Pan
Information System & Security and Countermeasures Experiment Center, Beijing Institute of Technology
1 shared publication- Ming Zhou
Langboat Technology
1 shared publication Senlin Luo
Information System & Security and Countermeasures Experiment Center, Beijing Institute of Technology
1 shared publication
Latest Publications
- 10 December 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Audio, Speech and Language Processing
- Vol. 34, 125-137
Emotion-Aware LLM Adaptation for Empathetic Dialogue Generation
The increasing emphasis on spiritual well-being in contemporary society has fueled a growing demand for empathetic dialogue generation. While large language models (LLMs) have demonstrated remarkable performance in dialogue generation tasks, their empathetic capabilities remain limited due to the absence of explicit emotional guidance. To address this limitation, we propose EmoDiag, an emotion-aware LLM adaptation framework designed to enhance the empathetic abilities of LLMs through fine-grained emotional control and a non-intrusive adaptation mechanism using the residual memory transformer (RMT). Specifically, EmoDiag employs RMT's encoder to predict the target emotional state from past conversations, which serves as an explicit control condition. Then, EmoDiag seamlessly integrates the predicted emotional state into the LLM's generation process through RMT's decoder at the logit level, enabling emotion-aware dialogue generation in a plug-and-play manner. Experimental results demonstrate that EmoDiag significantly outperforms baseline models in empathetic dialogue generation tasks and enhances the empathetic capacity of LLMs, while maintaining high-quality text generation in terms of diversity, fluency, and relevance.
- 10 December 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Audio, Speech and Language Processing
- Vol. 34, 125-137
Emotion-Aware LLM Adaptation for Empathetic Dialogue Generation
The increasing emphasis on spiritual well-being in contemporary society has fueled a growing demand for empathetic dialogue generation. While large language models (LLMs) have demonstrated remarkable performance in dialogue generation tasks, their empathetic capabilities remain limited due to the absence of explicit emotional guidance. To address this limitation, we propose EmoDiag, an emotion-aware LLM adaptation framework designed to enhance the empathetic abilities of LLMs through fine-grained emotional control and a non-intrusive adaptation mechanism using the residual memory transformer (RMT). Specifically, EmoDiag employs RMT's encoder to predict the target emotional state from past conversations, which serves as an explicit control condition. Then, EmoDiag seamlessly integrates the predicted emotional state into the LLM's generation process through RMT's decoder at the logit level, enabling emotion-aware dialogue generation in a plug-and-play manner. Experimental results demonstrate that EmoDiag significantly outperforms baseline models in empathetic dialogue generation tasks and enhances the empathetic capacity of LLMs, while maintaining high-quality text generation in terms of diversity, fluency, and relevance.
- Top Cited PapersOpen Access
- 6 October 2023
- Published by Association for Computing Machinery (ACM) in ACM Computing Surveys
- Vol. 56 (3) , 1-37
A Survey of Controllable Text Generation Using Transformer-based Pre-trained Language Models
Controllable Text Generation (CTG) is an emerging area in the field of natural language generation (NLG). It is regarded as crucial for the development of advanced text generation technologies that better meet the specific constraints in practical applications. In recent years, methods using large-scale pre-trained language models (PLMs), in particular the widely used Transformer-based PLMs, have become a new paradigm of NLG, allowing generation of more diverse and fluent text. However, due to the limited level of interpretability of deep neural networks, the controllability of these methods needs to be guaranteed. To this end, controllable text generation using Transformer-based PLMs has become a rapidly growing yet challenging new research hotspot. A diverse range of approaches have emerged in the past 3 to 4 years, targeting different CTG tasks that require different types of controlled constraints. In this article, we present a systematic critical review on the common tasks, main approaches, and evaluation methods in this area. Finally, we discuss the challenges that the field is facing, and put forward various promising future directions. To the best of our knowledge, this is the first survey article to summarize the state-of-the-art CTG techniques from the perspective of Transformer-based PLMs. We hope it can help researchers and practitioners in the related fields to quickly track the academic and technological frontier, providing them with a landscape of the area and a roadmap for future research.
- 6 October 2023
- Published by Association for Computing Machinery (ACM) in ACM Computing Surveys
- Vol. 56 (3) , 1-37
Top Cited PapersOpen AccessA Survey of Controllable Text Generation Using Transformer-based Pre-trained Language Models
Controllable Text Generation (CTG) is an emerging area in the field of natural language generation (NLG). It is regarded as crucial for the development of advanced text generation technologies that better meet the specific constraints in practical applications. In recent years, methods using large-scale pre-trained language models (PLMs), in particular the widely used Transformer-based PLMs, have become a new paradigm of NLG, allowing generation of more diverse and fluent text. However, due to the limited level of interpretability of deep neural networks, the controllability of these methods needs to be guaranteed. To this end, controllable text generation using Transformer-based PLMs has become a rapidly growing yet challenging new research hotspot. A diverse range of approaches have emerged in the past 3 to 4 years, targeting different CTG tasks that require different types of controlled constraints. In this article, we present a systematic critical review on the common tasks, main approaches, and evaluation methods in this area. Finally, we discuss the challenges that the field is facing, and put forward various promising future directions. To the best of our knowledge, this is the first survey article to summarize the state-of-the-art CTG techniques from the perspective of Transformer-based PLMs. We hope it can help researchers and practitioners in the related fields to quickly track the academic and technological frontier, providing them with a landscape of the area and a roadmap for future research.
- 10 October 2022
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Affective Computing
- Vol. 13 (4) , 1879-1891
Towards Contrastive Context-Aware Conversational Emotion Recognition
Conversational Emotion Recognition (CER) aims at classifying the emotion of each utterance in a conversation. For a target utterance, its emotion is jointly determined by multiple factors, such as conversation topics, emotion labels and intra/inter-speaker influences, in the conversational context of it. Then an important research question arises: can the effects of these contextual factors be sufficiently captured by the current CER models? To answer this question, we carry out an empirical study on four representative CER models by a context-replacement methodology. The results suggest that these models either exhibit a label-copying effect, or rely heavily on the intra/inter-speaker dependency structure within the conversation, but do not make a good use of the semantics carried by the conversational context. Thus, there is a high risk that they overfit certain single factors, yet lacking a holistic understanding of the semantic context. To tackle the problem, we propose a semantic-guided contrastive context-aware CER method, namely C3ER, to augment/regularize a backbone CER model, which can be any neural CER framework. Specifically, C3ER takes the hidden states of utterances from the CER model as input, extracts the contrast pairs consisting of relevant and irrelevant utterances to the conversational context of a target utterance, and uses contrastive learning to establish a soft semantic constraint between the target utterance and its context. It is then jointly trained with the main CER model, forcing the model to gain a semantic understanding of the context. Extensive experimental results show that C3ER can significantly boost the accuracy and improve the robustness of the representative CER models.
- 10 October 2022
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Affective Computing
- Vol. 13 (4) , 1879-1891
Towards Contrastive Context-Aware Conversational Emotion Recognition
Conversational Emotion Recognition (CER) aims at classifying the emotion of each utterance in a conversation. For a target utterance, its emotion is jointly determined by multiple factors, such as conversation topics, emotion labels and intra/inter-speaker influences, in the conversational context of it. Then an important research question arises: can the effects of these contextual factors be sufficiently captured by the current CER models? To answer this question, we carry out an empirical study on four representative CER models by a context-replacement methodology. The results suggest that these models either exhibit a label-copying effect, or rely heavily on the intra/inter-speaker dependency structure within the conversation, but do not make a good use of the semantics carried by the conversational context. Thus, there is a high risk that they overfit certain single factors, yet lacking a holistic understanding of the semantic context. To tackle the problem, we propose a semantic-guided contrastive context-aware CER method, namely C3ER, to augment/regularize a backbone CER model, which can be any neural CER framework. Specifically, C3ER takes the hidden states of utterances from the CER model as input, extracts the contrast pairs consisting of relevant and irrelevant utterances to the conversational context of a target utterance, and uses contrastive learning to establish a soft semantic constraint between the target utterance and its context. It is then jointly trained with the main CER model, forcing the model to gain a semantic understanding of the context. Extensive experimental results show that C3ER can significantly boost the accuracy and improve the robustness of the representative CER models.
- Open Access
- 29 May 2019
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 7, 69246-69256
An Efficient Android Malware Detection System Based on Method-Level Behavioral Semantic Analysis
According to the recent report, 12,000 new Android malware samples will be generated every day. Efficient identification of evolving malware is an urgent challenge. Traditional methods based on structured features such as permissions and sensitive Application Programming Interface(API) calls lack high-level behavioral semantics to detect evolving malware. The methods based on Call Graphs(CG) are good at behavioral semantic analysis but face the problem of huge time and space consumption, which leads to low detection efficiency. In this paper, we propose a novel Android malware detection method based on method-level correlation relationship of application’s abstracted API calls. Firstly, we split each Android application’s source code into separate function methods and just keep the abstracted API calls of them to form a set of abstracted API calls transactions. And then, we calculate the confidence of association rules between the abstracted API calls, which forms behavioral semantics to describe an application. Finally, we combine machine learning to identify the different behavioral patterns of malicious and benign apps to build the detection system. The results of our empirical evaluation show our system is competitive in terms of classification accuracy and detection efficiency. At dataset Drebin(benign 5.9K and malware 5.6K) and AMD(benign 20.5K and malware 20.8K), our system has achieved 96% and 98% detection results both in Accuracy and F-measure. Compared to state-of-the-art system in detecting evolving malware called MaMaDroid on dataset of 6.0K benign and 20.5K malicious samples spanning from 2010 to 2017, our system achieves higher accuracy while improving detection efficiency by 15 times (2.9s vs 45.7s per sample).
- 29 May 2019
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 7, 69246-69256
Open AccessAn Efficient Android Malware Detection System Based on Method-Level Behavioral Semantic Analysis
According to the recent report, 12,000 new Android malware samples will be generated every day. Efficient identification of evolving malware is an urgent challenge. Traditional methods based on structured features such as permissions and sensitive Application Programming Interface(API) calls lack high-level behavioral semantics to detect evolving malware. The methods based on Call Graphs(CG) are good at behavioral semantic analysis but face the problem of huge time and space consumption, which leads to low detection efficiency. In this paper, we propose a novel Android malware detection method based on method-level correlation relationship of application’s abstracted API calls. Firstly, we split each Android application’s source code into separate function methods and just keep the abstracted API calls of them to form a set of abstracted API calls transactions. And then, we calculate the confidence of association rules between the abstracted API calls, which forms behavioral semantics to describe an application. Finally, we combine machine learning to identify the different behavioral patterns of malicious and benign apps to build the detection system. The results of our empirical evaluation show our system is competitive in terms of classification accuracy and detection efficiency. At dataset Drebin(benign 5.9K and malware 5.6K) and AMD(benign 20.5K and malware 20.8K), our system has achieved 96% and 98% detection results both in Accuracy and F-measure. Compared to state-of-the-art system in detecting evolving malware called MaMaDroid on dataset of 6.0K benign and 20.5K malicious samples spanning from 2010 to 2017, our system achieves higher accuracy while improving detection efficiency by 15 times (2.9s vs 45.7s per sample).