A5100321165- Topic Modeling
- Natural Language Processing Techniques
- Advanced Graph Neural Networks
- Data Quality and Management
- Blockchain Technology Applications and Security
- Complex Systems and Time Series Analysis
- Anomaly Detection Techniques and Applications
- Advanced Text Analysis Techniques
- Quantum Information and Cryptography
- Semantic Web and Ontologies
- Domain Adaptation and Few-Shot Learning
- Human Pose and Action Recognition
- Multimodal Machine Learning Applications
- Economic theories and models
- Complex Network Analysis Techniques
- Text Readability and Simplification
- Speech and dialogue systems
- Magneto-Optical Properties and Applications
- Biomedical Text Mining and Ontologies
- Privacy-Preserving Technologies in Data
- Software-Defined Networks and 5G
- Quantum and electron transport phenomena
- Machine Learning and Algorithms
- Game Theory and Applications
- Data Stream Mining Techniques
Beijing Information Science & Technology University
2024
University of Illinois Urbana-Champaign
2021-2023
Zhengzhou University of Light Industry
2022
Hebei University of Technology
2021
Shanghai Jiao Tong University
2019-2021
Hohai University
2021
University of Pennsylvania
2020
PLA Army Engineering University
2019
The tasks of Rich Semantic Parsing, such as Abstract Meaning Representation (AMR), share similar goals with Information Extraction (IE) to convert natural language texts into structured semantic representations. To take advantage similarity, we propose a novel AMR-guided framework for joint information extraction discover entities, relations, and events the help pre-trained AMR parser. Our consists two components: 1) an based graph aggregator let candidate entity event trigger nodes collect...
Abstract Blockchain networks have attracted tremendous attention for creating cryptocurrencies and decentralized economies built on peer-to-peer protocols. However, the complex nature of dynamics feedback mechanisms within these economic has rendered it difficult to reason about growth evolution networks. Hence, proper mathematical frameworks model analyze behavior blockchain-enabled are essential. To address this need, we establish a formal framework, based dynamical systems, core concepts...
Xinya Du, Zixuan Zhang, Sha Li, Pengfei Yu, Hongwei Wang, Tuan Lai, Xudong Lin, Ziqi Iris Liu, Ben Zhou, Haoyang Wen, Manling Darryl Hannan, Jie Lei, Hyounghun Kim, Rotem Dror, Haoyu Michael Regan, Qi Zeng, Qing Lyu, Charles Carl Edwards, Xiaomeng Jin, Yizhu Jiao, Ghazaleh Kazeminejad, Zhenhailong Chris Callison-Burch, Mohit Bansal, Vondrick, Jiawei Han, Dan Roth, Shih-Fu Chang, Martha Palmer, Heng Ji. Proceedings of the 2022 Conference North American Chapter Association for Computational...
Decentralized Ledger Technology, popularized by the Bitcoin network, aims to keep track of a ledger valid transactions between agents virtual economy without central institution for coordination. In order faithful and accurate list transactions, is broadcast replicated across machines in peer-to-peer network. To enforce validity (i.e., no negative balance or double spending), network as whole coordinates accept reject new based on set rules aiming detect block operations malicious Byzantine...
Supervised event extraction models require a substantial amount of training data to perform well.However, annotation requires lot human effort and costs much time, which limits the application existing supervised approaches new types.In order reduce manual labor shorten time build an system for arbitrary ontology, we present framework train such systems more efficiently without large annotations.Our trigger labeling model uses weak supervision approach, only set keywords, small number...
Knowledge graph embedding is aimed at capturing the semantic information of entities by modeling structural between entities. For long-tail which lack sufficient information, general knowledge models often show relatively low performance in link prediction. In order to solve such problems, this paper proposes a framework learn as well attribute simultaneously. Under framework, H-AKRL (Hypergraph Neural Networks based Attribute-embodied Representation Learning) model put forward, where...
The emerging paradigm of Network Function Virtualization (NFV) promises to shorten the renewal cycles network functions and reduce capital expenses by flexibly deploying virtualized (VNFs) implementation on commodity servers. However, required resource each type (CPU, memory, etc.) for running VNF should be provisioned guarantee performance when processing packets. This comes with different deployment cost, especially in a heterogeneous cloud consisting large number function platforms from...
Cryptocurrencies and blockchain networks have attracted tremendous attention from their volatile price movements the promise of decentralization. However, most projects run on business narratives with no way to test verify assumptions promises about future. The complex nature system dynamics within networked economies has rendered it difficult reason growth evolution these networks. This paper drew concepts differential games, classical control engineering, stochastic dynamical come up a...
We target modeling latent dynamics in high-dimension marked event sequences without any prior knowledge about marker relations. Such problem has been rarely studied by previous works which would have fundamental difficulty to handle the arisen challenges: 1) high-dimensional markers and unknown relation network among them pose intractable obstacles for dynamic process; 2) one observed sequence may concurrently contain several different chains of interdependent events; 3) it is hard well...
Sensor-based human activity recognition (HAR) is a fundamental problem that can have broad impact on many research/industrial fields. The deep learning methods pave the way for extracting robust and informative features from non-stationary HAR data, achieving high-accuracy HAR. Most of works in literature consider closed set recognition, which assumes all classes activities are known both training test stages. However, it challenging to apply models real-world applications, contain unseen...
Fine-grained few-shot entity extraction in the chemical domain faces two unique challenges. First, compared with tasks general domain, sentences from papers usually contain more entities. Moreover, models have difficulty extracting entities of long-tailed types. In this paper, we propose Chem-FINESE, a novel sequence-to-sequence (seq2seq) based approach, to address these Our Chem-FINESE has components: seq2seq extractor extract named input sentence and self-validation module reconstruct...
Open-domain Question Answering (OpenQA) aims at answering factual questions with an external large-scale knowledge corpus. However, real-world is not static; it updates and evolves continually. Such a dynamic characteristic of poses vital challenge for these models, as the trained models need to constantly adapt latest information make sure that answers remain accurate. In addition, still unclear how well OpenQA model can transfer completely new domains. this paper, we investigate...
The dynamic nature of real-world information necessitates efficient knowledge editing (KE) in large language models (LLMs) for updating. However, current KE approaches, which typically operate on (subject, relation, object) triples, ignore the contextual and relation among different knowledge. Such methods could thus encounter an uncertain boundary, leaving a lot relevant ambiguity: Queries that be answered pre-edit cannot reliably afterward. In this work, we analyze issue by introducing...
The intelligent question answering over the knowledge graph aims to automatically answer natural language questions via locating correct entities in graph. Aside from former progresses, it is still challenging multi-relation because of variety and complexity language, as well combinatorial explosion on possible candidates. In this paper, we propose a novel embedding-based approach named SPE-QA address these issues. It answers by identifying its most semantic like question-answer path...
In order to solve the problem of lack effective methods for ontology inconsistency, a user preferences-oriented alignment repair model is proposed. This uses 0-1 linear programming method minimize remove cost; structure and source are used measure axiom importance; Finally, by choosing minimal conflict sets preferences limit strategy, purpose eliminating reducing semantic loss guaranteeing credibility achieved. The experimental results show that this can effectively restored more suitable...
Relations in most of the traditional knowledge graphs (KGs) only reflect static and factual connections, but fail to represent dynamic activities state changes about entities. In this paper, we emphasize importance incorporating events KG representation learning, propose an event-enhanced embedding model EventKE. Specifically, given original KG, first incorporate event nodes by building a heterogeneous network, where entity are distributed on two sides network inter-connected argument links....
Modern large-scale Pre-trained Language Models (PLMs) have achieved tremendous success on a wide range of downstream tasks. However, most the LM pre-training objectives only focus text reconstruction, but not sought to learn latent-level interpretable representations sentences. In this paper, we manage push language models obtain deeper understanding sentences by proposing new objective, Sparse Latent Typing, which enables model sparsely extract sentence-level keywords with diverse latent...