Hybrid Search for Korean Policy Data — pgvector plus Full-Text Signals
Why GRAXEL combines vector search and text search for Korean public-policy retrieval.
Korean policy search is hard because user questions and government wording rarely match exactly. A person may ask about rent support, while the official title uses administrative language.
Why this matters for GRAXEL
Pure keyword search misses semantic matches, while pure vector search can return plausible but legally irrelevant records. For government benefits, both precision and explanation matter.
GRAXEL combines vector similarity with full-text signals and then ranks results in a way that can be explained. The AI answer should not replace the source record; it should help users understand why a record appears.
Operational notes
- Keep the user-facing promise narrow enough that the service can be verified in a browser.
- Document the boundary between automated AI output and source-backed data so reviewers can understand the workflow.
- Link the implementation back to the public trust pages: About GRAXEL, Contact, and the platform overview.
For a small SaaS portfolio, trust comes from showing the real operating system behind the product: what runs, why it exists, and how it is maintained.
What changed in practice
Hybrid search makes MyHyetaek more trustworthy and turns policy data into a real service rather than a simple list of scraped pages. The same pattern now influences how the portal presents public services: planned ideas stay out of the main catalog, while usable beta services and documented operating notes receive stronger internal links.
When this article is read together with the monorepo operations note and the zero-cost infrastructure note, it gives a more complete view of how GRAXEL turns small service ideas into maintained products.
Official references
Share
Related articles
Continue with GRAXEL posts connected by topic and tags.
Operating MyHyetaek RAG — Making 11,600 Government Policies Searchable
How GRAXEL structures policy data, hybrid search, and AI responses for the MyHyetaek government-benefits assistant.
Running Korean, English, and Japanese Pages with Next.js 15 and next-intl
How GRAXEL handles locale-specific routing, metadata, and crawler-facing content across ko, en, and ja.
Using Ollama for Local LLM Workflows in a Production-Oriented System
Where local LLMs fit inside GRAXEL and where managed APIs are still the safer choice.