{"id":16713,"date":"2024-12-20T20:22:41","date_gmt":"2024-12-21T04:22:41","guid":{"rendered":"https:\/\/www.couchbase.com\/blog\/?p=16713"},"modified":"2025-06-13T16:36:25","modified_gmt":"2025-06-13T23:36:25","slug":"embedding-models","status":"publish","type":"post","link":"https:\/\/www.couchbase.com\/blog\/pt\/embedding-models\/","title":{"rendered":"O que s\u00e3o modelos de incorpora\u00e7\u00e3o? Uma vis\u00e3o geral"},"content":{"rendered":"

What are embedding models?<\/span><\/h2>\n

Embedding models are a type of machine learning model designed to represent data (such as text, images, or other forms of information) in a continuous, low-dimensional vector space. These <\/span>embeddings<\/span><\/a> capture semantic or contextual similarities between pieces of data, enabling machines to perform tasks like comparison, clustering, or classification more effectively.<\/span><\/p>\n

Imagine you want to describe different fruits. Instead of long descriptions, you use numbers for characteristics like sweetness, size, and color. For example, an apple might be [8, 5, 7] while a banana is [9, 7, 4]. These numbers make it easier to compare or group similar fruits.<\/span><\/p>\n

What does an embedding model do?<\/span><\/h2>\n

An embedding model converts text, images, and audio into meaningful numbers and compares them to find patterns or connections. This process is similar to how a library organizes books by genre or topic, allowing users to find what they’re looking for faster.<\/span><\/p>\n

Here are examples of daily use cases for embedding models:<\/span><\/p>\n

Text search<\/b><\/p>\n

Imagine typing “best Greek food” into a search engine. An embedding model will convert your query into numbers and retrieve documents with similar embeddings. The model will then show results that are close to your query.<\/span><\/p>\n

Recommend movies<\/b><\/p>\n

If you liked a movie, the system uses an embedding model to represent it (e.g., genre, cast, mood) as numbers. It compares these numbers to other movie embeddings and recommends similar ones.<\/span><\/p>\n

Match images and captions<\/b><\/p>\n

An embedding model can match an image of a sunset over the ocean with the caption “A serene sunset over calm ocean waves” by converting both the image and potential captions into numerical representations (embeddings). The model identifies the caption with an embedding closest to the image\u2019s embedding, ensuring an accurate match. This technique powers tools like image search and photo tagging.<\/span><\/p>\n

Group similar items<\/b><\/p>\n

A shopping website uses embeddings to group similar products together. For instance, “red sneakers” might be close to “blue sneakers” in the embedding space, so they\u2019re shown as related.<\/span><\/p>\n

Types of embeddings models<\/span><\/h2>\n

There are several embedding models, each designed for different types of data and tasks. Here are the main types:<\/span><\/p>\n

Word embedding models<\/span><\/h3>\n

These models convert words into numerical vectors that capture semantic meanings and relationships between words. Examples include:<\/span><\/p>\n