~/projects/aboutlk-knowledge-graphs ← projects

/projects$cat aboutlk-knowledge-graphs.html

About.lk · Tourism Engine & Knowledge Graphs

iLabs · 2025 · applied R&D

the problem

A tourism app engine is only as good as its knowledge. The raw material was huge volumes of unstructured text — places, routes, facts, reviews — none of it queryable in a way that supports accurate, relationship-aware answers. Plain vector RAG flattens that structure; tourism is full of structure (this town is near that beach, known for that food).

what i built

I ran the applied R&D to convert that unstructured data into read/write knowledge graphs in Neo4j, and built a continuous update pipeline so the graph keeps growing as new data arrives. To do it well I implemented techniques from recent research — Socratic KG for question-driven extraction and GraphRAG-R1 for graph-grounded retrieval — enabling highly accurate semantic retrieval and automated data extraction.

Compared to plain vector search, the graph captures how entities relate, not just which ones are similar — which matters a lot for a domain built on connections.