# From Words to Coordinates — Embeddings & Vector Search

Source: https://www.geekswithgeeks.com/en/embeddings/b-idea

> Understand the central idea: similar meaning, nearby vectors.

## Closeness stands for similarity

An **embedding** is a list of numbers (a **vector**) that represents an item, such as a sentence, a document, an image or a product, so that **items with similar meaning get vectors that are close together** and unrelated items get vectors that are far apart. Computers cannot compare the *meaning* of two sentences directly, but they can compare two lists of numbers quickly. Once text is embedded, tasks such as **semantic search** ("find passages about holidays even if they say vacation"), **recommendation**, **clustering**, **duplicate detection**, **classification** and **retrieval for LLMs (RAG)** become geometry: find the nearest vectors. Typical text embeddings have a few hundred to a few thousand dimensions.

## Meaning as a point in space

An embedding maps text, images or other items to a vector so that similar items sit close together.

![Four ideas: vector, distance, model, dimensions.](assets/figures/embeddings/section-1-map.svg) — Figure 1.1 — Vector, distance, model and dimensions.

## A map of ideas

On a city map, nearby places are close in real life. An embedding is a map of meaning: sentences about leave land in one neighbourhood and sentences about travel in another.

## Embeddings are not the text

A vector cannot be turned back into the exact text. Always store the original text (or a pointer to it) next to its vector.

**Quiz:** What should be true of the embeddings of two sentences with similar meaning?

- [ ] Their vectors are far apart
- [ ] Their vectors are identical
- [x] Their vectors are close together
- [ ] They have the same length in characters

*Answer:* Their vectors are close together. Embedding models are trained so that similar meaning gives nearby vectors.