Resources

We are committed to promoting open research in the field of knowledge intelligence, sharing our research projects, datasets, benchmarks, and tools with the academic and industrial communities.

Research Projects #

Our research focuses on cutting-edge fields including trustworthy artificial intelligence, intelligent information retrieval, natural language processing, and knowledge intelligence applications, committed to promoting the deep integration of theoretical innovation and practical applications.

AI Security and Trust #

In today’s rapidly evolving AI landscape, ensuring the security and trustworthiness of AI systems is crucial. We conduct in-depth research in this frontier field:

Core Research Areas:

• Trustworthy AI Algorithms: Design AI algorithms with robustness, explainability, and fairness
• AIGC Security: Research on security risks and protection mechanisms for AI-generated content
• AI Model Reliability: Enhance the stability of deep learning models under adversarial attacks
• Privacy-Preserving AI: Develop federated learning and differential privacy technologies to protect user privacy

Intelligent Information Retrieval #

Information retrieval is the key technology connecting user needs with massive information. We focus on personalized and intelligent information retrieval system research:

Core Research Areas:

• Recommendation Systems: Deep learning-based personalized recommendation algorithms
• User Modeling and Understanding: Multi-dimensional user profiling and behavior prediction
• Knowledge Acquisition: Automatic acquisition of structured knowledge from multi-source heterogeneous data
• Search Technologies: Next-generation intelligent search engine technologies

Natural Language Processing #

Natural language processing is the core technology for achieving intelligent human-machine interaction. We conduct cutting-edge research in large-scale language models and multimodal understanding:

Core Research Areas:

• Large Language Models (LLM): Optimization and application of pre-trained models
• Multimodal Knowledge Extraction: Knowledge extraction fusing visual and linguistic information
• Knowledge Representation Learning: Vectorized representation and reasoning of textual knowledge
• Dialogue Systems: Knowledge-driven intelligent dialogue technologies

Spatial-Temporal Analysis and Prediction #

Spatial-temporal data analysis has important application value in intelligent transportation, urban computing, and other fields:

Core Research Areas:

• Location-aware Social Network Analysis: Geographic location-based social behavior modeling
• Real-time Push Algorithms: Spatial-temporal aware information push systems
• Traffic Data Analysis: Data mining and prediction for intelligent transportation systems
• Time Series Prediction: Deep learning prediction models for multivariate time series

Knowledge Intelligence Applications #

Apply knowledge intelligence technologies to real industry scenarios, solving complex real-world problems:

Core Research Areas:

• Intelligent Finance: Knowledge graph-based financial risk control and investment decision-making
• Intelligent Transportation: Traffic flow optimization and intelligent navigation systems
• Recommendation System Applications: Industrial applications in e-commerce, content recommendation, and other fields
• Multimedia Content Understanding: Intelligent analysis and understanding of videos and images

Datasets #

Knowledge Graph Datasets #

CN-Academic-KG #

Chinese Academic Knowledge Graph

A large-scale knowledge graph containing over 1 million Chinese academic entities and relationships, covering multiple disciplines including computer science, mathematics, and physics.

GitHub Paper

MultiModal-KG #

Multimodal Knowledge Graph

A multimodal knowledge graph that integrates textual, visual, and numerical information, containing over 500K entities and 2M+ triples.

GitHub Paper

Natural Language Processing Datasets #

FewShot-RE-ZH #

Chinese Few-Shot Relation Extraction

A dataset specifically designed for Chinese few-shot relation extraction, containing 20 relation categories with only 1-5 annotated samples per category.

GitHub Paper

Dialogue-KG #

Dialogue Knowledge Graph

A knowledge graph built from real dialogues, used to evaluate the knowledge reasoning capabilities of dialogue systems.

GitHub Paper

Benchmarks #

KG-Benchmark Suite #

We have developed a comprehensive knowledge graph benchmark suite covering multiple core tasks:

Entity Linking #

• Dataset: 5 domains, 100K+ annotated samples
• Evaluation Metrics: Precision, Recall, F1-Score
• Baseline Models: 10 SOTA model comparisons

Relation Extraction #

• Dataset: Bilingual (Chinese-English), 50 relation types
• Evaluation Metrics: Micro/Macro F1, AUC
• Baseline Models: 15 classic and state-of-the-art models

Knowledge Reasoning #

• Dataset: Multi-hop reasoning, complex queries
• Evaluation Metrics: MRR, Hits@K
• Baseline Models: 8 graph neural network models

GitHub Online Evaluation

Open Source Tools #

KG-Toolkit #

# Installation
pip install kg-toolkit

# Usage Example
from kg_toolkit import KnowledgeGraph, EntityExtractor

# Create knowledge graph
kg = KnowledgeGraph()

# Entity extraction
extractor = EntityExtractor(model='bert-base-chinese')
entities = extractor.extract("Zhang San is a professor at Sun Yat-sen University")

Key Features:

• 🚀 Multiple pre-trained model support
• 📊 Knowledge graph visualization
• 🔧 Flexible API interface
• 📚 Detailed documentation and tutorials

GitHub PyPI

NLP-Utils #

Main Modules:

• Chinese text segmentation and POS tagging
• Named entity recognition
• Relation extraction
• Text classification
• Semantic similarity computation

Tutorials and Documentation #

Getting Started Tutorials #

Knowledge Graph Fundamentals #

A beginner-friendly knowledge graph tutorial covering theoretical foundations and practical operations. Start Learning

Graph Neural Networks in Practice #

An accessible introduction to the principles and implementation of graph neural networks. Start Learning

NLP Pipeline #

A complete NLP project tutorial from data preprocessing to model deployment. Start Learning

Technical Blog #

Regularly updated technical blog sharing latest research findings and technical insights:

• Applications of Knowledge Graphs in Recommender Systems
• Recent Advances in Few-Shot Learning
• Development Trends in Multimodal AI