Semantic Search

Semantic search is a retrieval approach that matches the intent and meaning of a query instead of only its literal words. Text is converted to embeddings, and the system returns documents whose vectors are closest to the query vector.

Why it matters

Keyword search misses relevant results when wording differs: a search for "site not loading" will not find a page titled "troubleshooting connection timeouts" unless the words overlap. Semantic search bridges that gap and powers help-centre search, product discovery and retrieval for AI assistants. It can also surface loosely related or wrong results, and it handles exact identifiers, error codes and rare names poorly, so many systems combine it with keyword search in a hybrid approach.

Example

Query: "website keeps going offline"
Keyword match: needs those words
Semantic match: "Diagnosing intermittent server downtime" (score 0.82)

Common mistakes

• Dropping keyword search entirely, which hurts queries with exact SKUs, IDs or quoted phrases.
• Skipping evaluation. Test with real queries and judged results.
• Embedding very long documents whole instead of splitting them into passages.

Related

• Embedding
• Cosine similarity
• Vector database
• RAG

यह पेज सामान्य जानकारी है, जिसे हमारी संपादकीय टीम ने लिखा है और दस्तावेज़ीकरण बदलने पर अपडेट किया जाता है। सर्वर और प्रदाता अलग-अलग होते हैं, इसलिए बदलाव पहले सुरक्षित तरीके से जाँचें। देखें: संपादकीय नीति और अस्वीकरण.