Document embeddings are one of the simplest ways to give machines an understanding of text, and in our previous article, Document Embeddings Explained: A Guide for Beginners, we explored how they turn entire documents into dense numerical vectors that capture meaning and context. Now that you understand what embeddings are and why they’re useful for tasks like semantic search, classification, and clustering, this tutorial will show you how to generate them in practice using Python. Whether you’re working with short paragraphs, long articles, or a collection of documents, the steps in this guide will help you create embeddings that you can analyze, compare, and use in real applications.

This guide will walk you through the process of generating document embeddings using Python and a pretrained transformer model. Even if you’ve never worked with embeddings before, you’ll see how easy it is to convert text into meaningful numerical vectors that can be used for search, similarity comparison, clustering, and many other AI applications.

Objectives

By the end of this article, you will be able to:

• Understand how to generate document embeddings using Python.
• Install and use the SentenceTransformers library to create text embeddings.
• Load a pretrained transformer model suitable for document-level representation.
• Convert sentences, paragraphs, or full documents into numerical vector formats.
• Compute semantic similarity between documents using cosine similarity.
• Apply embeddings to simple, practical tasks such as comparing text similarity.

Prerequisites

Before we dive into the guide on how to generate simple document embeddings with Python, make sure you have the following ready:

• We will be using Google Colab as our development environment for this guide. If you don’t know how to set it up, you can follow our previous article guide: Data Preprocessing Guide for Beginners in ML.
• You need to install the necessary libraries for this guide. Running this command on the Google Colab cell will install them.

!pip install sentence-transformers scikit-learn

Generating the Document Embeddings

Importing the libraries

After installing, you can immediately import the libraries in the next cell. 

Preparing your data

The next step now that the model is loaded and working is to make a simple sentence dataset that we can practice on. You can make your own list of sentences or copy the following code.

# Example documents
documents = [
"Artificial intelligence is transforming businesses.",
"Machine learning allows computers to learn from data.",
"Python is a popular programming language for AI applications.",
"Deep learning models require large amounts of data.",
"AI can help automate repetitive tasks and improve efficiency."
]

Generate the embeddings

With our simple dataset, we can now start generating their document embeddings with this code. This code will first encode the data using the pretrained model and print out the shape of the generated embeddings.

# Generate embeddings for all documents
embeddings = model.encode(documents)
# Check the shape of the embeddings array
print("Embeddings shape:", embeddings.shape)