Advances in Knowledge Graph Computation in the Era of Generative AI

Join us in exploring knowledge graph computation in the era of generative AI

The goal of this tutorial talk is to motivate and give a comprehensive introduction to the essential concepts of graph-based knowledge representation and construction, methods in knowledge graph computation, higher-order knowledge relations with hypergraph, and roles of knowledge graphs in the wave of generative AI enhanced applications and beyond. Knowledge graphs (KGs) provide a framework for organizing, accessing, and deriving knowledge through the interplay of symbolic representing and deductive reasoning, and statistical (inductive) learning. KGs empower a wide range of intelligent applications across domains like healthcare, finance, transportation, education, service industry, and various science and technology domains.

The intended audience in this talk includes researchers and practitioners who are new to KGs or are interested in incorporating KGs based approaches to leveraging domain specific applications from various perspectives. As such, we do not assume that audiences have specific expertise on knowledge graphs.

Knowledge graph computation is a big topic in terms of the scopes of the domain dependent depth and breath. This talk focuses on some limited and fundamental issues along with briefly introducing the basic concepts in knowledge graphs.

The talk focuses on the following research questions:

• How can semantic completeness and consistency be ensured within KGs?
• How to extract and build KGs from raw data?
• How to interplay symbolic representing and deductive reasoning with statistical (inductive) learning through the upper stream of KGs construction and down stream of KGs applications?

The talk will address the following issues and topics:

1. Foundation of graph-based knowledge representation and construction
   1. Modelling and representing KGs
      1. RDF (Resource Description Framework) and RDF schema
      2. Structural KGs vs. attributed KGs
      3. Actional knowledge graphs
      4. Temporal knowledge graphs
   2. KGs construction
      1. Explicit KGs (deductive KGs)
      2. Implicit KGs (inductive KGs)
   3. KGs completion
   4. KGs validation and measurement
      1. Structural quality
      2. Knowledge quality
      3. Entropy for measuring knowledge graphs
2. Methods in knowledge graph computation
   1. Explicit KGs computation
      1. Rule-based approaches
      2. Ontology-driven methods
      3. Schema matching and integration
      4. Crowdsourced/manual curation
   2. Implicit KGs computation
      1. KGs embedding
      2. BERT-based models and ChatGPT-4 to learn KGs from text, etc.
   3. Latent knowledge induction
      1. Matrix and tensor factorization, topic modeling, and clustering
   4. Relation extraction via deep learning
      1. Transformer-based models and sequence labeling
3. Higher-order knowledge relations with hypergraph
   1. Concepts of knowledge hypergraph (KHGs) computation
   2. Methods for KHGs construction
   3. Approaches to KHGs computation
   4. Applications of KHGs
      1. KHGs augmented cognitive intelligence
      2. KHGs enhanced representation for learning experience
4. KGs enhanced generative AI applications