Research

We are developing efficient data mining algorithms that enable fast and memory-optimized analysis of large-scale datasets. Our research focuses on subgraph matching and subgraph query processing, which facilitate the discovery of complex patterns within large graph structures. We also explore advanced string pattern matching techniques for identifying sequences in massive text data. We also aim to design approximate nearest neighbor search algorithms that provide rapid similarity retrieval in high-dimensional spaces, supporting applications utilizing multi-modal data. Our goal is to create scalable solutions that accelerate data-driven insights while minimizing computational resources. Our recent research topics are:

• Efficient graph mining algorithms

• Efficient retrieval algorithms

• Efficient pattern matching algorithms

We are dedicated to advancing techniques for uncovering hidden patterns and insights within vast knowledge graphs and knowledge bases. Our research mainly focuses on knowledge-intensive tasks, such as knowledge graph completion enabling the inference of missing relationships between entities, and link prediction which identifies potential connections within complex networks. We develop cutting-edge systems for knowledge graph question answering (KGQA), empowering machines to provide accurate and contextually relevant answers. Our work extends to multi-hop question answering, where models reason over multiple relational paths to derive comprehensive answers, pushing the boundaries of machine reasoning and retrieval-augmented generation (RAG). Our recent research topics are:

• Multi-hop question answering (in extensive use or specifically for computational social science)

• Knowledge graph question answering

• Knowledge graph completion

• Link prediction

We are developing intelligent recommender systems that deliver personalized suggestions by analyzing user preferences and behavioral patterns. Our research explores sequential recommendation, capturing temporal dynamics to predict future interactions, and content-based filtering, which tailors recommendations based on item characteristics. We also investigate collaborative filtering techniques that leverage collective user data to enhance prediction accuracy, with applications ranging from e-commerce and entertainment to healthcare. Our ongoing research topics are:

• Text-based recommendation

• Sequential recommendation

• Next-POI recommendation

• Medication recommendation