Generate Gremlin queries using Amazon Bedrock models

This article discusses a novel approach to generating Gremlin queries using Amazon Bedrock models, specifically addressing the complexity of querying graph databases for non-technical users.

• Developed a three-step methodology for converting natural language to Gremlin queries
• Integrated graph knowledge and domain-specific context to improve query accuracy
• Used LLM-based evaluation to assess query generation and execution
• Tested the approach across 120 questions with two models
• Achieved an overall query generation accuracy of 74.17%

The solution aims to help business analysts and data scientists interact with graph databases more easily by translating natural language queries into technical Gremlin queries, with promising results in accuracy and efficiency.