A5006839003- Natural Language Processing Techniques
- Topic Modeling
- Advanced Welding Techniques Analysis
- Biomedical Text Mining and Ontologies
- Aluminum Alloys Composites Properties
- Welding Techniques and Residual Stresses
- Text Readability and Simplification
- Privacy-Preserving Technologies in Data
- Computational and Text Analysis Methods
- Advanced Computing and Algorithms
- Authorship Attribution and Profiling
- Color perception and design
- Power Systems Fault Detection
- Hand Gesture Recognition Systems
- Educational Technology and Pedagogy
- MXene and MAX Phase Materials
- Sunflower and Safflower Cultivation
- Powdery Mildew Fungal Diseases
- HVDC Systems and Fault Protection
- Hearing Impairment and Communication
- Web Data Mining and Analysis
- Sentiment Analysis and Opinion Mining
- Stochastic Gradient Optimization Techniques
- Speech and dialogue systems
- Geographic Information Systems Studies
Beihang University
2021-2024
Dalian University of Technology
2022-2024
Center for Outcomes Research and Clinical Epidemiology
2024
Institute of Art
2024
Shanghai University of Electric Power
2023
Minzu University of China
2022
Xiamen University
2020-2021
Abstract Federated learning (FL) is a novel distributed machine paradigm that enables participants to collaboratively train centralized model with privacy preservation by eliminating the requirement of data sharing. In practice, FL often involves multiple and requires third party aggregate global information guide update target participant. Therefore, many methods do not work well due training test each participant may be sampled from same feature space underlying distribution. Meanwhile,...
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...
In recent years, for the structural characteristics and design requirements of integral rotor disc shaft integrated engine, welding quality mechanical properties superalloy weldments have received increasing attention. this paper, inertia friction (IFW) FGH96 alloy was carried out using different parameters, homogeneous connection hollow bars successfully realized. The microstructure evolution, fracture failure welded joints at room high temperatures were investigated. IFW divided into weld...
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,...
Research on document-level Neural Machine Translation (NMT) models has attracted increasing attention in recent years. Although the proposed works have proved that inter-sentence information is helpful for improving performance of NMT models, what should be regarded as context remains ambiguous. To solve this problem, we a novel cache-based model which conducts dynamic caching guided by theme-rheme information. The experiments NIST evaluation sets demonstrate our achieves substantial...
The prior knowledge, such as expert rules and knowledge base, has been proven effective in the traditional Biomedical Named Entity Recognition (BioNER). Most current neural BioNER systems use this external for pre-processing or post-editing instead of incorporate it into training process, which cannot be learned by model. To encode model, we present a unified multi-task Machine Reading Comprehension (MRC) framework BioNER. Specifically, MRC task, question sequences are derived from standard...
Biomedical Named Entity Recognition (BioNER) aims to identify biomedical domain-specific entities (e.g. gene, chemical and disease) from unstructured texts. Despite deep learning-based methods for BioNER achieving satisfactory results, there is still much room improvement. Firstly, most existing use independent sentences as training units ignore inter-sentence context, which usually leads the labeling inconsistency problem. Secondly, previous document-level works have approved that...
6061-T6 aluminum matrix composites reinforced by AlCoCrFeNi2.1 high-entropy alloy (HEA) were obtained using friction stir processing (FSP), and the resultant further strengthened T6 heat treatment. The effects of reinforcement phase content on microstructure mechanical properties investigated. results showed that HEA particles homogeneously distributed in under effect multi-pass reciprocating processing. particle-stimulated nucleation mechanism decreased composite grain size with increase...
Inertia friction welding (IFW) was used to join large-diameter hollow bars made of Inconel 690 and 316LN successfully. The interfacial characteristics, microstructure, mechanical properties fracture mechanism welded joints under different process parameters were investigated. results indicated that a joining with interlocking metallurgical bonding found in IFW joints. There significant mixing zone at the interface. elemental diffusion layer “wrinkles” zone. A tiny quantity C elements...
Federated learning (FL) is a novel distributed machine paradigm that enables participants to collaboratively train centralized model with privacy preservation by eliminating the requirement of data sharing. In practice, FL often involves multiple and requires third party aggregate global information guide update target participant. Therefore, many methods do not work well due training test each participant may be sampled from same feature space underlying distribution. Meanwhile, differences...
In recent years, natural language processing (NLP) models have demonstrated remarkable performance in text classification tasks. However, trust the decision-making process requires a deeper understanding of operational principles these networks. Therefore, there is an urgent need to enhance transparency and interpretability "black boxes". Aligned with this, we propose model-agnostic method named MCG. This generates counterfactual interpretations that are more faithful original models'...
The de facto way of utilizing black-box large language models (LLMs) to perform various downstream tasks is prompting. However, obtaining suitable prompts for specific still a challenging problem. While existing LLM-based methods demonstrate promising performance in task-oriented dialogue (TOD) task, they often require manual adjustment prompt selection, or focus solely on understanding generation. To address these issues, we propose an adaptive generation framework fully unleash the...