Semantic search

Definition

Semantic search uses vector embeddings to measure content similarity between a query and a set of documents. Unlike classical full-text search (BM25, TF-IDF), it also finds hits without a literal keyword match — synonyms, paraphrases, or cross-lingual equivalents, for example. Hybrid approaches combine semantic and lexical search to optimize recall and precision at the same time.

How Swiss Knowledge Hub uses this term

Swiss Knowledge Hub uses semantic search as the foundation of the retrieval step and complements it with keyword filters (for example case numbers or contract IDs) and reranking when needed, so domain-specific questions are answered precisely.

Related terms

• Embedding
• Vector database (vector store)
• RAG (Retrieval-Augmented Generation)
• Reranking

Sources

1. Wikipedia: Semantic search — https://en.wikipedia.org/wiki/Semantic_search