"Ask-A-Graph"
KGC 2026
•
27m
https://drive.google.com/file/d/1tH57rvRUGo-pWjm09u5AtXnARkPLrWxw/view?usp=share_link
Abstract
The gap between domain experts who need knowledge graph insights and the technical Cypher expertise required to query them represents one of the most significant adoption barriers in enterprise graph databases. While recent advances in large language models have enabled natural language interfaces, production systems face unique challenges: biomedical entity disambiguation, schema complexity, injection safety, and maintaining query correctness across heterogeneous graph topologies.
We present Ask-A-Graph, a production-grade natural language to Cypher pipeline that synthesizes techniques from the best available published research into a cohesive, enterprise-ready architecture. Our hybrid approach combines:
The system implements a novel two-tool disambiguation pattern that separates candidate identification from selection application, enabling stateful multi-turn conversations while maintaining lean orchestrator context. Built on the Model Context Protocol (MCP) and A2A agent standards, Ask-A-Graph demonstrates how thoughtful architecture can transform research prototypes into production systems serving pharmaceutical scientists exploring drug-disease-gene relationships.
This talk will share our architectural decisions, research synthesis methodology, and lessons learned from deploying natural language graph interfaces in regulated biomedical environments.
Up Next in KGC 2026
-
Building Real-Time Intelligence Syste...
https://drive.google.com/file/d/1GqheySdjCqY7N0bYpxm1bz495V1KYG1_/view?usp=share_link
Event-driven systems generate rich signals, but without context they remain fragmented and hard to reason over. Knowledge graphs offer a powerful way to connect these signals — yet building and operating them i...
-
Lessons Learned from 25 Years in Sema...
"https://drive.google.com/file/d/1A4hiMKMIX6x1lrAintov1-Kf5ion4xcb/view?usp=share_link
Knowledge Graphs Are Not Enough. In this session, we’ll challenge the conventional wisdom around knowledge graphs. While they’ve been positioned as the cornerstone of enterprise semantics, they are not the end...
-
Taxonomy Design Best Practices
"https://drive.google.com/file/d/1x8vYa4d8CmunWXKRcCX1Exm6SczzRUe9/view?usp=share_link
Ontologies form the semantic framework for knowledge graphs, but to serve the purpose of connecting to content and standardizing metadata, ontologies need to be based on taxonomies and other controlled vocabul...
