A5021246376- Topic Modeling
- Natural Language Processing Techniques
- Speech Recognition and Synthesis
- Advanced Text Analysis Techniques
- Web Data Mining and Analysis
- Biomedical Text Mining and Ontologies
- Water Quality Monitoring and Analysis
- Advanced Computational Techniques and Applications
- Speech and dialogue systems
- Multimodal Machine Learning Applications
- Text Readability and Simplification
- Hand Gesture Recognition Systems
- Hearing Impairment and Communication
- Human Pose and Action Recognition
- Scientific Computing and Data Management
- Music and Audio Processing
- Speech and Audio Processing
- Text and Document Classification Technologies
- Constructed Wetlands for Wastewater Treatment
- Expert finding and Q&A systems
- Liver Disease Diagnosis and Treatment
- Service-Oriented Architecture and Web Services
- Rough Sets and Fuzzy Logic
- Algorithms and Data Compression
- Organ Transplantation Techniques and Outcomes
Xiamen University
2016-2025
Nanjing University of Aeronautics and Astronautics
2024-2025
State Ethnic Affairs Commission
2025
North Minzu University
2025
Institute of Electronics
2025
Ningxia University
2025
Zhejiang University
2024
Chinese Academy of Cultural Heritage
2023
Zhengzhou University
2023
South China University of Technology
2021
Semantic Role Labeling (SRL) is believed to be a crucial step towards natural language understanding and has been widely studied. Recent years, end-to-end SRL with recurrent neural networks (RNN) gained increasing attention. However, it remains major challenge for RNNs handle structural information long range dependencies. In this paper, we present simple effective architecture which aims address these problems. Our model based on self-attention can directly capture the relationships between...
Given the advance of Internet technologies, we can now easily extract hundreds or thousands news stories any ongoing incidents from newswires such as CNN.com, but volume information is too large for us to capture blueprint. Information retrieval techniques topic detection and tracking are able organize events, in a flat hierarchical structure, within topic. However, they incapable presenting complex evolution relationships between events. We interested learn not only what major events also...
Neural machine translation (NMT) heavily relies on word-level modelling to learn semantic representations of input sentences. However, for languages without natural word delimiters (e.g., Chinese) where sentences have be tokenized first, conventional NMT is confronted with two issues: 1) it difficult find an optimal tokenization granularity source sentence modelling, and 2) errors in 1-best tokenizations may propagate the encoder NMT. To handle these issues, we propose word-lattice based...
Cross-domain sentiment analysis has achieved promising results with the help of pre-trained language models. As GPT-3 appears, prompt tuning been widely explored to enable better semantic modeling in many natural processing tasks. However, directly using a fixed predefined template for cross-domain research cannot model different distributions \operatorname{[MASK]} token domains, thus making underuse technique. In this paper, we propose novel Adversarial Soft Prompt Tuning method (AdSPT)...
Neural machine translation (NMT) heavily relies on word-level modelling to learn semantic representations of input sentences.However, for languages without natural word delimiters (e.g., Chinese) where sentences have be tokenized first,conventional NMT is confronted with two issues:1) it difficult find an optimal tokenization granularity source sentence modelling, and2) errors in 1-best tokenizations may propagate the encoder NMT.To handle these issues, we propose word-lattice based...
Knowledge representation is one of the critical problems in knowledge engineering and artificial intelligence, while embedding as a methodology indicates entities relations graph low-dimensional, continuous vectors. In this way, compatible with numerical machine learning models. Major methods employ geometric translation to design score function, which weak-semantic for natural language processing. To overcome disadvantage, paper, we propose our model based on multi-view clustering...
Sign language translation (SLT) is an important application to bridge the communication gap between deaf and hearing people. In recent years, research on SLT based neural frameworks has attracted wide attention. Despite progress, current still in initial stage. fact, systems perform poorly processing long sign sentences, which often involve long-distance dependencies require large resource consumption. To tackle this problem, we propose two explainable adaptations traditional models using...
Sign Language Translation (SLT) is a promising technology to bridge the communication gap between deaf and hearing people. Recently, researchers have adopted Neural Machine (NMT) methods, which usually require large-scale corpus for training, achieve SLT. However, publicly available SLT very limited, causes collapse of token representations inaccuracy generated tokens. To alleviate this issue, we propose Con-SLT, novel token-level Contrastive learning framework , learns effective by...
In the transitional region between agriculture and livestock rearing in northern China, planting forage crops rows among fruit trees as feed orchards represents an effective strategy for enhancing ecological environment while addressing increasing demand feed. Nonetheless, impact of short-term mowing cover on quality soil remains unclear. A two-year field experiment was conducted Ziziphus jujuba cv. “Lingwuchangzao” located Lingwu County, Ningxia Hui Autonomous Region, arid semi-arid...
With the rapid development of UAV technology, delivery has gained attention for its potential to reduce labor costs. However, limitations in load capacity and energy restrict UAVs’ distribution capabilities. This paper proposes a cooperative scheme combining traditional trucks UAVs extend coverage improve completion rates. For densely distributed depots wide-area regions, we develop algorithms task allocation path planning truck-independent system. Specifically, minimum-cost, maximum-flow...
Introduction Tetraena mongolica was established in the West Ordos Region of northwest China approximately 140 million years ago. It plays an irreplaceable role maintaining local ecosystem stability. Methods This study aimed to evaluate effects planting T. on soil nutrition and microbial communities by comparing root zone (Rz_soil) bare (B_soil) across three different plant communitie. Results The results showed that decreased pH Na + while increasing available potassium, organic matter,...
Objectives As more families participate expanded newborn screening for metabolic disorders in China, the overall number of false positives increases. Our goal was to assess potential impact on parental stress, perceptions child's health, and family relationships. Methods Parents 49 infants with false-positive results panel were compared parents 42 children normal results. first completed structured interview using likert scales, closed open questions. also parenting stress index. Results A...
Neural biomedical named entity recognition (BioNER) methods usually require a large amount of annotated data, while the BioNER datasets are often difficult to obtain and small in scale due limitations privacy, ethics high degree specialization.To alleviate lack training samples, unlike conventional that only use token-level information, this paper proposes method simultaneously utilize latent multi-granularity information dataset.Concretely, proposed model is based on multi-task approach,...
Few-shot continual relation extraction aims to continually train a model on incrementally few-shot data learn new relations while avoiding forgetting old ones. However, current memory-based methods are prone overfitting memory samples, resulting in insufficient activation of and limited ability handle the confusion similar classes. In this paper, we design N-way-K-shot Continual Relation Extraction (NK-CRE) task propose novel method with Consistent Prototype Learning (ConPL) address...