Semantic Search
Semantic search uses AI embeddings to understand the meaning behind a reader's query, not just the exact words they type. This enables Vellumine to return relevant results even when the query and the content use completely different wording.
How It Works
Traditional keyword search matches the literal terms in a query against the words in your content. Semantic search goes further by converting both queries and content into numerical representations (embeddings) that capture meaning. This enables:
- Conceptual matching -- A search for "how to get more readers" can match a post titled "Growing Your Blog Audience", even though none of the words overlap.
- Natural language queries -- Readers can search conversationally (e.g. "what's the best way to optimize images") and still find relevant results.
- Cross-language potential -- Meaning-based matching can surface relevant content even when the query language differs slightly from the content language.
Hybrid Search
When semantic search is enabled, Vellumine uses a hybrid approach that combines both keyword matching and semantic matching. This means you get the best of both worlds:
- Exact keyword matches are still prioritized when they exist.
- Semantic understanding fills in the gaps when exact matches don't exist.
- Results are ranked by a combination of keyword relevance and semantic similarity.
Enabling Semantic Search
- Navigate to the Search Configuration page.
- Toggle Semantic Search on.
- Vellumine will generate embeddings for your content in the background. This may take a moment depending on the size of your index.
Once enabled, semantic search applies to all searches on your blog automatically. No changes to your script tag or integration are needed.
When to Use Semantic Search
Semantic search is most valuable when:
- Your blog covers broad topics where readers may use varied terminology.
- You notice a high zero-results rate despite having relevant content.
- Your audience tends to search using natural language or questions rather than specific keywords.
- Your content uses technical terminology that readers may not know.