Embedding

An embedding is a technique that turns the "meaning" of words into numerical coordinates. It's like placing every sentence on a map: the more similar two sentences are in meaning, the closer their coordinates sit. "Dog treat recommendations" and "pet supplements" barely share any words, but in embedding coordinates, they're neighbors.

This matters because computers can finally compare things "by meaning, not by words." Old-style search needed matching keywords, but embedding-based search (semantic search) finds results that mean the same thing even when the wording is completely different.

It's the most widely used component under the hood of AI products. Search, recommendations ("articles similar to this one"), and the "retrieve relevant documents" step in RAG all run on embeddings. When an article mentions a "vector database" or "semantic search," that's embeddings at work.

"The new embedding model improved search accuracy" — this means the "meaning coordinates" used for meaning-based search were made more precise.

1. Try sending this to a chatbot: "Group these five sentences by similar meaning: ①Recommend dog treats ②The stock market crashed ③What are good supplements for dogs? ④The KOSPI fell sharply ⑤Will it rain tomorrow?"
2. Sentences ①③ and ②④ get grouped together — even though they barely share any words. They were grouped by "distance in meaning," not by shared words, and embeddings are what measure that distance.
3. The "related articles" section at the bottom of METAL LAB articles works the same way — embeddings find pieces that connect in content even when the wording differs.