A5090714679- Topic Modeling
- Natural Language Processing Techniques
- Text and Document Classification Technologies
- Advanced Text Analysis Techniques
- Network Security and Intrusion Detection
- Network Packet Processing and Optimization
- Internet Traffic Analysis and Secure E-voting
- Speech Recognition and Synthesis
- Software Engineering Research
- Advanced Computational Techniques and Applications
- Advanced Malware Detection Techniques
- Web Data Mining and Analysis
- Anomaly Detection Techniques and Applications
- Sentiment Analysis and Opinion Mining
- Spam and Phishing Detection
- Advanced Graph Neural Networks
- Data Quality and Management
Institute of Information Engineering
2023-2025
Chinese Academy of Sciences
2023-2025
University of Chinese Academy of Sciences
2023-2024
Mining structured knowledge from tweets using named entity recognition (NER) can be beneficial for many downstream applications such as recommendation and intention under standing. With tweet posts tending to multimodal, multimodal (MNER) has attracted more attention. In this paper, we propose a novel approach, which dynamically align the image text sequence achieve multi-level cross-modal learning augment textual word representation MNER improvement. To specific, our framework split into...
Multimodal fusion aims to improve the performance of models for applications by extracting and fusing information in different modalities, including texts, images or others. Recent researches have shown that multimodal is beneficial many multimedia tasks. In this paper, we study typical classification tasks social media posts, sarcasm detection sentiment analysis. This paper proposes DMF-RHGT-HPA, dynamic Fusion fusion(DMF), a relation-aware heterogeneous graph transformer(RHGT) hierarchical...
Document-level relation extraction (DocRE) aims to extract relations between entities from unstructured document text. Compared sentence-level extraction, it requires more complex semantic understanding a broader text context. Currently, some studies are utilizing logical rules within evidence sentences enhance the performance of DocRE. However, in data without provided sentences, researchers often obtain list for entire through retrieval (ER). Therefore, DocRE suffers two challenges:...
In recent years, text classification methods based on neural networks and pre-trained models have gained increasing attention demonstrated excellent performance. However, these still some limitations in practical applications: (1) They typically focus only the matching similarity between sentences. there exists implicit high-value information both within sentences of same class across different classes, which is very crucial for tasks. (2) Existing such as language graph-based approaches...
Large language models (LLMs) are widely applied in various fields of society due to their powerful reasoning, understanding, and generation capabilities. However, the security issues associated with these becoming increasingly severe. Jailbreaking attacks, as an important method for detecting vulnerabilities LLMs, have been explored by researchers who attempt induce generate harmful content through attack methods. Nevertheless, existing jailbreaking methods face numerous limitations, such...
Recognizing the users of devices (or clusters devices) who use IP addresses as unique identities on Internet can easily enable numerous security applications. Fast and accurate user recognition is critical for supervisors to find influenced organizations connected their networks in light new threats. Many users’ information scatters multisource data addresses. Up until now, has had two main problems. On one hand, existing methods could not fully wastes valuable labels. other only a tiny...