Knowledge Graph Application Cases: Meituan Brain, Sage Knowledge Base, and Other Industry Scenarios

Knowledge graphs are special graph structures that combine semantic and topological information and have been widely adopted in recommendation, healthcare, and many other industry scenarios.

Case 01 – Meituan Brain (Lifestyle Services) – Since 2018 Meituan’s NLP Center has built a massive knowledge graph to improve merchant and user experiences. Four major challenges (explainability, domain diversity, data sparsity, and spatio‑temporal complexity) are addressed through:

Graph‑structured information display (e.g., showing drug effects or product filters based on graph data).

Using graph paths to guide recommendations, such as recalling related queries when a user’s original query yields no clicks.

Two recall methods: direct graph‑path recall (e.g., recommending "bubble milk tea" for the query "milk tea") and embedding‑based recall using GNN‑trained embeddings.

Graph‑based reasoning to generate recommendation explanations (e.g., "Users from the same hometown also like this restaurant").

Case 02 – Fourth Paradigm Sage Knowledge Base – A low‑threshold, end‑to‑end knowledge‑graph platform supporting QA, recommendation, drug discovery, and stock prediction. The article outlines representation‑learning techniques from triples to sub‑graphs:

Triple‑based models : translation‑based (TransE), MLP, ConvE, RSN, and bilinear models (AutoSF, AutoSF+ with progressive and genetic‑algorithm search).

Relation‑path models : extending triples to multi‑hop paths (PTransE) and recurrent skipping networks (RSN) for long‑term dependencies.

Graph Neural Network models : R‑GCN, CompGCN, KE‑GCN, GraIL (inductive sub‑graph reasoning), and RED‑GNN (dynamic‑programming‑based sub‑graph aggregation with attention).

Case 03 – Other Knowledge‑Graph Applications – Examples include logic‑rule reasoning (RNNLogic) at Mila AI Lab and a power‑grid knowledge graph from China Electric Power Research Institute.

All cases are extracted from the "AI Technology Application Cases Manual" (knowledge‑graph module, pages 201 and 140) and can be accessed via the provided QR codes.