Embedding
In short: What is Embedding?
An embedding is the conversion of text or other data into a list of numbers (a vector) such that items with similar meaning end up close together.
Explanation
A dedicated embedding model maps each piece of text to a vector made up of hundreds or thousands of numbers. Semantic similarity between two passages can then be measured by the distance between their vectors, regardless of exact word overlap.
This powers semantic search, clustering of related content and recommendations. The retrieval step in RAG pipelines is built on embeddings as well.
Example: for the query “how do I cancel my subscription?”, the system finds the section titled “Terminating your agreement” even though the two share no words in common.