A5017147857- Topic Modeling
- Natural Language Processing Techniques
- Text and Document Classification Technologies
- scientometrics and bibliometrics research
- Complex Network Analysis Techniques
- Advanced Graph Neural Networks
- Multimodal Machine Learning Applications
- Speech and dialogue systems
- Biomedical Text Mining and Ontologies
- Advanced Text Analysis Techniques
- Anomaly Detection Techniques and Applications
- Machine Learning in Healthcare
- Opinion Dynamics and Social Influence
- Sentiment Analysis and Opinion Mining
- Recommender Systems and Techniques
- Domain Adaptation and Few-Shot Learning
- Innovation and Knowledge Management
- Career Development and Diversity
- Advanced Bandit Algorithms Research
- Speech and Audio Processing
- Intelligent Tutoring Systems and Adaptive Learning
- Catalysts for Methane Reforming
- Web Data Mining and Analysis
- Machine Learning and Data Classification
- Online Learning and Analytics
University of Hong Kong
2021-2025
Harbin Engineering University
2024
First Hospital of Shanxi Medical University
2024
University of Illinois Urbana-Champaign
2024
Shanxi Medical University
2023-2024
China University of Geosciences
2024
Amazon (United States)
2019-2023
Amazon (Germany)
2019-2023
Second Hospital of Shanxi Medical University
2023
Hunan University
2023
Lacking of adaptation to various array imperfections is an open problem for most high-precision direction-of-arrival (DOA) estimation methods. Machine learning-based methods are data-driven, they do not rely on prior assumptions about geometries, and expected adapt better when compared with model-based counterparts. This paper introduces a framework the deep neural network address DOA problem, so as obtain good enhanced generalization unseen scenarios. The consists multitask autoencoder...
Being able to recognize words as slots and detect the intent of an utterance has been a keen issue in natural language understanding. The existing works either treat slot filling detection separately pipeline manner, or adopt joint models which sequentially label while summarizing utterance-level without explicitly preserving hierarchical relationship among words, slots, intents. To exploit semantic hierarchy for effective modeling, we propose capsule-based neural network model accomplishes...
User intent detection plays a critical role in question-answering and dialog systems. Most previous works treat as classification problem where utterances are labeled with predefined intents. However, it is labor-intensive time-consuming to label users’ intents diversely expressed novel will continually be involved. Instead, we study the zero-shot problem, which aims detect emerging user no currently available. We propose two capsule-based architectures: IntentCapsNet that extracts semantic...
Scientific collaboration is essential in solving problems and breeding innovation. Coauthor network analysis has been utilized to study scholars' collaborations for a long time, but these studies have not simultaneously taken different features into consideration. In this paper, we present systematic approach analyze the differences possibilities that two authors will cooperate as seen from effects of homophily, transitivity, preferential attachment. Exponential random graph models (ERGMs)...
Open relation extraction is the task of extracting open-domain facts from natural language sentences. Existing works either utilize heuristics or distant-supervised annotations to train a supervised classifier over pre-defined relations, adopt unsupervised methods with additional assumptions that have less discriminative power. In this work, we propose self-supervised framework named SelfORE, which exploits weak, signals by leveraging large pretrained model for adaptive clustering on...
Network embedding aims at projecting the network data into a low-dimensional feature space, where nodes are represented as unique vector and structure can be effectively preserved. In recent years, more online application service sites massive complex networks, which extremely challenging for traditional machine learning algorithms to deal with. Effective of low-dimension representation both save storage space enable applicable handle data. performance will degrade greatly if networks sparse...
To alleviate human efforts from obtaining large-scale annotations, Semi-Supervised Relation Extraction methods aim to leverage unlabeled data in addition learning limited samples. Existing self-training suffer the gradual drift problem, where noisy pseudo labels on are incorporated during training. noise labels, we propose a method called MetaSRE, Label Generation Network generates accurate quality assessment by (meta) successful and failed attempts Classification as an additional...
With the large language model showing human-like logical reasoning and understanding ability, whether agents based on can simulate interaction behavior of real users, so as to build a reliable virtual recommendation A/B test scene help application research is an urgent, important economic value problem. The combination design machine learning provide more efficient personalized user experience for products services. This service meet specific needs users improve satisfaction loyalty. Second,...
Text queries are naturally encoded with user intentions. An intention detection task tries to model and discover intentions that in text queries. Unlike conventional classification tasks where the label of is highly correlated some topic-specific words, words from different topic categories tend co-occur medical related Besides existence word order, correlations way organized into sentence crucial tasks.
Low-resource Relation Extraction (LRE) aims to extract relation facts from limited labeled corpora when human annotation is scarce. Existing works either utilize self-training scheme generate pseudo labels that will cause the gradual drift problem, or leverage meta-learning which does not solicit feedback explicitly. To alleviate selection bias due lack of loops in existing LRE learning paradigms, we developed a Gradient Imitation Reinforcement Learning method encourage label data imitate...
Abstract Background IPF is a complex lung disease whose aetiology not fully understood, but diet may have an impact on its development and progression. Therefore, we investigated the potential causal connection between dietary intake through TSMR to offer insights for early prevention recommendations. Methods The study incorporated 29 exposure factors, oily fish intake, bacon processed meat poultry beef pork lamb/mutton non-oily fresh fruit cooked vegetable baked bean tomato tinned salad/raw...
Abstract With the increasing demand for ocean resource exploitation, deep-sea exploration, and environmental protection, importance of underwater operation technologies has become more prominent. Traditional rigid robotic arms lack flexibility in complex environments, limiting their effectiveness diverse tasks. Flexible arms, with pliable structures, offer superior adaptability. However, influence geometric design on performance water-driven actuators remains unclear, hindering optimization....
In this paper, we formulate a more realistic and difficult problem setup for the intent detection task in natural language understanding, namely Generalized Few-Shot Intent Detection (GFSID). GFSID aims to discriminate joint label space consisting of both existing intents which have enough labeled data novel only few examples each class. To approach problem, propose model, Conditional Text Generation with BERT (CG-BERT). CG-BERT effectively leverages large pre-trained model generate text...
Shuliang Liu, Xuming Hu, Chenwei Zhang, Shu'ang Li, Lijie Wen, Philip Yu. Proceedings of the 2022 Conference North American Chapter Association for Computational Linguistics: Human Language Technologies. 2022.
Developing a coal to ethylene glycol (CtEG) process is of great interest many countries, especially China. However, because the hydrogen carbon ratio coal-gasified gas far less than desired value, CtEG suffers from high CO2 emission and wastes precious resources. At same, most coke oven (COG) discharged directly or used as fuel, resulting in waste resources, serious environmental pollution, economic loss. To develop efficient clean utilization COG we propose novel assisted (CaCtEG) process....