Building Next‑Generation Data Intelligence Infrastructure with Knowledge Graphs: From New Infrastructure to Cognitive AI

The talk introduces the concept of "new infrastructure" (新基建) in China, highlighting its seven pillars—5G, data centers, AI, industrial internet, high‑speed rail, ultra‑high voltage, and EV charging—while focusing on the data center and AI components.

It then describes the DIKW model (Data‑Information‑Knowledge‑Wisdom) and argues that knowledge graphs are the essential bridge from massive data to intelligent, cognitive applications.

AI is portrayed as evolving from operational (perception) intelligence to cognitive and eventually general intelligence; current AI remains at the perception layer, and knowledge graphs provide the reasoning and understanding needed for the next stage.

The article outlines the challenges of traditional data governance—low utilization of unstructured data, difficulty integrating multimodal data, poor relationship exploitation, lack of flexible business models, and insufficient intelligent application support.

To address these issues, a knowledge‑graph‑centric data governance framework is proposed, including a unified knowledge representation model, automated knowledge extraction from structured and unstructured sources, multi‑strategy information extraction, deep semantic fusion (ontology alignment, entity alignment, relationship discovery, entity linking), and polymorphic storage built around graph databases.

Based on this foundation, an intelligent data‑governance platform is built, offering semantic search, QA, recommendation, and other AI services.

The discussion extends to a cognitive‑intelligence middle‑platform powered by knowledge graphs, emphasizing micro‑service‑based components, pre‑built models, and business orchestration to enable rapid, plug‑and‑play AI applications.

Industrial practice examples include financial risk control, intelligence analysis, and domain‑specific bots for insurance, e‑commerce, and anti‑corruption, all leveraging the same knowledge‑graph‑driven middle‑platform.

Finally, the speaker, Dr. Hu Fanghuai, summarizes the session and encourages the audience to share, like, and follow the community.