Using Semantic Search Effectively

Keyword vs semantic

Papyrus runs both keyword (full-text) and semantic (vector) search, then blends results. You don't need to choose — but understanding the difference helps you craft better queries.

Keyword: matches the literal words in your query
Semantic: matches the meaning — finds documents that talk about the concept even if they don't use the exact words

Examples:

Query	Keyword finds	Semantic also finds
“contract termination”	docs with “contract termination”	docs about “ending an agreement”, “exit clause”
“leave policy”	docs with “leave policy”	docs about “annual leave”, “PTO”, “time off”
“kra pin”	docs with “kra pin”	docs about “tax registration”, “KRA compliance”

Writing better queries

Be specific about the document type: “invoice from Acme” beats just “Acme”
Add time anchors: “Q1 2026 expense reports” beats just “expense reports”
Use natural language: “policies about working from home” works as well as the right keywords
Combine filters: even semantic search benefits from a filter (“classification: contracts”, “tagged: 2025”) to narrow the corpus

Filters

Below the search bar, filter results by:

File type (PDF, Word, Excel, image)
Date created / modified
Folder
Author / owner
Tags
Classification
AI document type
Status (Active, Approved, Archived, etc.)

Filters compose. Use them aggressively when the result count is overwhelming.

Saving searches

If you find yourself running the same query repeatedly:

Run the query
Click Save as Smart Folder or Save search
Name it; it appears in your sidebar

The Smart Folder version updates continuously — new matching documents show up automatically.

Searches respect RBAC

You only see results for documents you have access to. If you search for something you'd expect to find but don't, ask an admin whether the document exists in a folder you can't see — most “missing” search results are permission, not indexing, issues.