A5003055263- Topic Modeling
- Natural Language Processing Techniques
- Speech and dialogue systems
- Advanced Text Analysis Techniques
- Soil Mechanics and Vehicle Dynamics
- Agricultural Engineering and Mechanization
- Power Line Communications and Noise
- Multimodal Machine Learning Applications
- Underwater Vehicles and Communication Systems
- Electrocatalysts for Energy Conversion
- Telecommunications and Broadcasting Technologies
- Intelligent Tutoring Systems and Adaptive Learning
- Digital Imaging for Blood Diseases
- HVDC Systems and Fault Protection
- Advanced MIMO Systems Optimization
- Machine Learning in Healthcare
- Fluid Dynamics and Vibration Analysis
- Retinal Imaging and Analysis
- Software-Defined Networks and 5G
- Retinal and Optic Conditions
- Maritime Navigation and Safety
- Ship Hydrodynamics and Maneuverability
- Advanced Photocatalysis Techniques
- Innovative Teaching and Learning Methods
- Explainable Artificial Intelligence (XAI)
University of Chinese Academy of Sciences
2023-2025
Beijing National Laboratory for Molecular Sciences
2024-2025
Chinese Academy of Sciences
2024-2025
Institute of Chemistry
2024-2025
Second Affiliated Hospital of Zhejiang University
2021-2025
North University of China
2025
University of Copenhagen
2025
Dalian University of Technology
2023-2024
Institute for Infocomm Research
2019-2024
Agency for Science, Technology and Research
2019-2024
Due to the lack of publicly available resources, conversation summarization has received far less attention than text summarization. As purpose conversations is exchange information between at least two interlocutors, key about a certain topic often scattered and spanned across multiple utterances turns from different speakers. This phenomenon more pronounced during spoken conversations, where speech characteristics such as backchanneling false-starts might interrupt topical flow. Moreover,...
Weak adhesion and lack of underwater self-healability hinder advancing soft iontronics particularly in wet environments like sweaty skin biological fluids. Mussel-inspired, liquid-free ionoelastomers are reported based on seminal thermal ring-opening polymerization a biomass molecule α-lipoic acid (LA), followed by sequentially incorporating dopamine methacrylamide as chain extender, N,N'-bis(acryloyl) cystamine, lithium bis(trifluoromethanesulphonyl) imide (LiTFSI). The exhibit universal to...
Summarizing conversations via neural approaches has been gaining research traction lately, yet it is still challenging to obtain practical solutions. Examples of such challenges include unstructured information exchange in dialogues, informal interactions between speakers, and dynamic role changes speakers as the dialogue evolves. Many result complex coreference links. Therefore, this work, we investigate different explicitly incorporate abstractive summarization models tackle aforementioned...
COPD poses a significant global public health challenge, primarily characterised by irreversible airflow restriction and persistent respiratory symptoms. The hallmark pathology of includes sustained airway inflammation the eventual destruction lung tissue structure. While multiple risk factors are implicated in disease's progression, underlying mechanisms remain largely elusive. perpetuation is pivotal to advancement COPD, emphasising importance investigating these self-sustaining for deeper...
In this paper, we propose a controllable neural generation framework that can flexibly guide dialogue summarization with personal named entity planning. The conditional sequences are modulated to decide what types of information or perspective focus on when forming summaries tackle the under-constrained problem in tasks. This supports two use cases: (1) Comprehensive Perspective, which is general-purpose case no user-preference specified, considering summary points from all conversational...
The internet is an abundant source of news every day. Thus, efficient algorithms to extract keywords from the text are important obtain information quickly. However, precision and recall mature keyword extraction need improvement. TextRank, which derived PageRank algorithm, uses word graphs spread weight words. propagation in TextRank focuses only on frequency. To improve performance we propose Semantic Clustering (SCTR), a semantic clustering algorithm based TextRank. Firstly, vectors...
Developing highly active electrocatalysts for overall water splitting is critical the extensive implementation of renewable hydrogen.
Annotated data plays a critical role in Natural Language Processing (NLP) training models and evaluating their performance. Given recent developments Large Models (LLMs), such as ChatGPT demonstrate zero-shot capability on many text-annotation tasks, comparable with or even exceeding human annotators. Such LLMs can serve alternatives for manual annotation, due to lower costs higher scalability. However, limited work has leveraged complementary annotators, nor explored how annotation is best...
Development of efficient and environmentally friendly energy conversion technology has become a top priority with the continuous improvement global demand environmental awareness. Designing an electrocatalyst high activity, excellent stability durability is major problem in development water decomposition. In this study, polymetallic sulfides FeS-Co3S4-Ni3S2/NF (shortened as FeCoNi-S/NF) nano-flower structure was synthesized by multi-step conventional hydrothermal method. FeCoNi-S/NF...
Abstract Eosinophils play a crucial role as effector cells in asthma pathogenesis, with their differentiation being tightly regulated by metabolic mechanisms. While the involvement of iron various cellular processes is well known, its specific eosinophil has largely remained unexplored. This study demonstrates that levels are increased during process from progenitors to mature and activated eosinophils context allergic airway inflammation. Through experiments involving chelators,...
Large language models (LLMs) have shown impressive multilingual capabilities through pretraining on diverse corpora. While these show strong reasoning abilities, their performance varies significantly across languages due to uneven training data distribution. Existing approaches using machine translation, and extensive cross-lingual tuning face scalability challenges often fail capture nuanced processes languages. In this paper, we introduce AdaCoT (Adaptive Chain-of-Thought), a framework...
Diabetes mellitus is a global health threat, with chronic wounds, including oral mucosal being severe complication. These wounds are characterized by delayed healing and increased inflammation due to hyperglycemia, affecting patients' quality of life. Current treatments for cannot offer sustained management these injuries in diabetic patients. Here, glucose/ROS dual-responsive hydrogel incorporating sitagliptin was developed the treatment wounds. After chemical modification tetra-armed...
Achieving robust adhesion of bioadhesives on wet tissues to block gastric perforation remains a challenge due the gradually deteriorated adhesive-tissue interactions by interfacial acidity and multienzyme fluids, thus accompanying failure shedding life-threatening risks. Here, we report biocompatible double-umbrella-shaped endoscopy-deliverable hydrogel occluder (EHO) made caffeic acid (CA)-grafted chitosan (CS) polyacrylamide (PAM) molding technique, which is capable customizable, rapid,...
To simulate the interactions between a pneumatic cotton precision seed-metering device and coated delinted seeds accurately, physical simulation experiments based on rotating drum apparatus were combined to calibrate discrete element parameters of E`kangmian-10 seeds. Firstly, contact dynamic repose angle measured through tests. Based particle size requirement Discrete Element Method (DEM) Computational Fluid Dynamics (CFD) coupling reverse engineering technology, seed bonded-particle model...
Few-shot fault diagnosis (i.e., with few samples) is a challenging issue in practice because samples are scarce and difficult to obtain. Data augmentation based on generative adversarial networks (GANs) has proven be an effective solution. However, it often encounters problems such as model training low quality of generated samples. In this article, improved GAN filtering mechanism developed for data augmentation, which introduces the self-attention instance normalization (IN) into structure...
We introduce AudioBench, a new benchmark designed to evaluate audio large language models (AudioLLMs). AudioBench encompasses 8 distinct tasks and 26 carefully selected or newly curated datasets, focusing on speech understanding, voice interpretation, scene understanding. Despite the rapid advancement of models, including multimodal versions, significant gap exists in comprehensive benchmarks for thoroughly evaluating their capabilities. addresses this by providing relevant datasets...