A5100348629- Topic Modeling
- Natural Language Processing Techniques
- Speech and dialogue systems
- Simulation and Modeling Applications
- Biomedical Text Mining and Ontologies
- Dam Engineering and Safety
- 3D Surveying and Cultural Heritage
- Advanced Decision-Making Techniques
- Multimodal Machine Learning Applications
- COVID-19 diagnosis using AI
- Advanced Graph Neural Networks
- Semantic Web and Ontologies
- Guidance and Control Systems
- Military Defense Systems Analysis
- Microgrid Control and Optimization
- Phonocardiography and Auscultation Techniques
- Social Robot Interaction and HRI
- Landslides and related hazards
- Text Readability and Simplification
- Educational Technology and Assessment
- Geotechnical Engineering and Analysis
- Advanced Neural Network Applications
- Data Stream Mining Techniques
- Satellite Image Processing and Photogrammetry
- Machine Learning in Healthcare
Hohai University
2014-2024
Dalian University of Technology
2023-2024
Institute of Information Engineering
2023
Chinese Academy of Sciences
2014-2023
Central China Normal University
2023
University of Chinese Academy of Sciences
2023
Ministry of Water Resources of the People's Republic of China
2022
NetEase (China)
2020
Zhejiang University
2019
Institute of Automation
2014
Most existing fine-tuned biomedical large language models (LLMs) focus on enhancing performance in monolingual question answering and conversation tasks. To investigate the effectiveness of LLMs diverse natural processing (NLP) tasks different languages, we present Taiyi, a bilingual LLM for NLP
Abstract Machine learning (ML) models have been developed to identify randomised controlled trials (RCTs) accelerate systematic reviews (SRs). However, their use has limited due concerns about performance and practical benefits. We a high-recall ensemble model using Cochrane RCT data enhance the identification of RCTs for rapid title abstract screening in SRs evaluated externally with our annotated datasets. Additionally, we assessed impact terms labour time savings recall improvement under...
This paper investigates a new task named Conversational Question Generation (CQG) which is to generate question based on passage and conversation history (i.e., previous turns of question-answer pairs). CQG crucial for developing intelligent agents that can drive question-answering style conversations or test user understanding given passage. Towards end, we propose approach Reinforced Dynamic Reasoning network, the general encoder-decoder framework but incorporates reasoning procedure in...
In recent years, with the rapid development of Internet technology and applications, scale data has exploded, which contains a significant amount valuable knowledge. The best methods for organization, expression, calculation, deep analysis this knowledge have attracted great deal attention. graph emerged as rich intuitive way to express Knowledge reasoning based on graphs is one current research hot spots in played an important role wireless communication networks, intelligent question...
Machine Comprehension (MC) is one of the core problems in natural language processing, requiring both understanding and knowledge about world. Rapid progress has been made since release several benchmark datasets, recently state-of-the-art models even surpass human performance on well-known SQuAD evaluation. In this paper, we transfer learned from machine comprehension to sequence-to-sequence tasks deepen text. We propose MacNet: a novel encoder-decoder supplementary architecture widely used...
There are usually a small amount and short sequence of prototype monitoring data on structural behavior during dangerous dam reinforcement. According to above characteristics, some methods, such as support vector machine, particle swarm optimization, genetic algorithm, rough set, combined build the real-time model safety status reinforcement dam. Firstly, construction principle standard machine-based is demonstrated. To improve modeling accuracy efficiency, optimization algorithm introduced...
How to accurately understand low-resource languages is the core of task-oriented human-computer dialogue system. Language understanding consists two sub-tasks, i.e., intent detection and slot filling. Intent still faces challenges due semantic ambiguity implicit intentions with users’ input. Moreover, separately modeling filling significantly decrease correctness relevance between questions answers. To address these issues, we propose a joint method using asynchronous training strategy. The...
Quiz question annotation aims to assign the most relevant knowledge point a question, which is key technology support intelligent education applications. However, existing methods only extract explicit semantic information that reveals literal meaning of and ignore implicit highlights intention. To this end, an innovative dual-channel model, Semantic-Knowledge Mapping Network (S-KMN) proposed enrich representation from two perspectives, knowledge, simultaneously. It integrates features...
Objective: Most existing fine-tuned biomedical large language models (LLMs) focus on enhancing performance in monolingual question answering and conversation tasks. To investigate the effectiveness of LLMs diverse NLP tasks different languages, We present Taiyi, a bilingual LLM for Materials Methods: first curated comprehensive collection 140 text mining datasets (102 English 38 Chinese datasets) across over 10 task types. Subsequently, two-stage strategy is proposed supervised fine-tuning...
Event argument extraction is critical to various natural language processing tasks for providing structured information. Existing works usually extract the event arguments one by one, and mostly neglect build dependency information among roles, especially from perspective of structure. Such an approach hinders model learning interactions between different roles. In this paper, we raise our research question: How adequately dependencies roles better performance? To end, propose intra-event...
This paper investigates a new task named Conversational Question Generation (CQG) which is to generate question based on passage and conversation history (i.e., previous turns of question-answer pairs). CQG crucial for developing intelligent agents that can drive question-answering style conversations or test user understanding given passage. Towards end, we propose approach Reinforced Dynamic Reasoning (ReDR) network, the general encoder-decoder framework but incorporates reasoning...
Profiling game players, especially potential churn and payment prediction, is of paramount importance for online games to improve the product design revenue. However, current solutions view either or prediction as an independent task most previous attempts only depend on single data source, i.e., tabular portrait data. Based two real-world games, we conduct extensive analysis. On one hand, there exists a significant correlation between player payment. other heterogeneous multi-source data,...
Applying Gaussian Splatting to perception tasks for 3D scene understanding is becoming increasingly popular. Most existing works primarily focus on rendering 2D feature maps from novel viewpoints, which leads an imprecise language field with outlier languages, ultimately failing align objects in space. By utilizing masked images extraction, these approaches also lack essential contextual information, leading inaccurate representation. To this end, we propose a Language-Embedded Surface Field...
Geological logging is a fundamental technical practice in coal mine and also one of the most important jobs for production mine, having significance mine’s safe source recycling. Irregular shapes excavating faces make current photographic geological mode methods unfeasible, so this paper studies critical techniques underground face mines proposes new close-range photogrammetry control method, orienting face’s stereoscopic image using directing laser plumb line, proves through experiments...
In a single-stage photovoltaic grid-connected control system based on quasi-Z source inverter (qZSI), the current inner loop controller improved active disturbance rejection (LADRC) is designed to improve robustness, and parameters of LADRC are optimized by sparrow search algorithm (SSA) further performance. The simulation results show that compared with traditional PI control, SSA optimization has superior
Indoor precise control field can be used for providing Object-space information in Close Range Photogrammetry. The technique and method to devise the which meet multipurpose of research needs fast,conveniently accurately is an important issue digital photogrammetry. paper discusses establish simulation design system indoor with VRML VB, uses guide construction field,achieves corresponding function. result shows that,on aspect three-dimensional visualization field, modeling approach flexible...
Air combat target tactical intent refers to the analysis and inference of enemy's intentions in real-time, adversarial environments by extracting battlefield environmental information, static attributes, real-time dynamic information air targets, combined with knowledge from military domain. To achieve this goal, many machine learning-based methods have been proposed infer aircraft intentions. However, these are only applicable individual cannot predict entire formation. Therefore, we...