TTT #5: Basics of Vector Search and Embeddings: A Practical Guide

Stephen CollinsAug 11, 2023

Vector search and embeddings have revolutionized the way we handle text data, driving advancements in Natural Language Processing (NLP), Machine Learning, and more. In this newsletter issue, I’ll introduce these topics and showcase their practical applications.

Understanding the Concept of Embeddings

Embeddings are multi-dimensional numerical representations of words, sentences, or documents. They capture the semantic meaning, preserving relationships and similarities.

Types of Embeddings

  • Word Embeddings: Represent individual words (e.g., Word2Vec, GloVe).
  • Sentence Embeddings: Represent entire sentences or paragraphs (e.g., SentenceTransformer).

Vector search allows you to find the nearest vectors corresponding to a particular query vector. It plays a critical role in tasks such as similarity search, recommendation systems, and more.

Exploring Cosine Similarity

Cosine similarity is a vital measure used in vector search. It calculates the cosine of the angle between two vectors, allowing us to gauge how similar they are.

Why Cosine Similarity?

  • Effectiveness: Robust measure for text similarity.
  • Scalability: Suitable for large-scale computations.
  • Interpretability: Intuitive understanding of similarity.

Practical Python Example: Leveraging Libraries

Follow a step-by-step Python tutorial from my blog post on getting started with vector search that demonstrates how to use the SentenceTransformer, sklearn, and the openai package.

Use Cases and Applications

Some use cases for vector search:

  • Search Engines: Enhance search relevance.
  • Recommendation Systems: Suggest related content.
  • Sentiment Analysis: Understand user sentiments.
  • Language Translation: Facilitate accurate translations.
  • Question & Answering Systems: Enables context-rich responses from LLMs by integrating context within user prompts.


Vector search and embeddings are fundamental to many applications in NLP and beyond. By understanding these concepts, you can build applications that can combine private or novel data with the capability of modern NLP models like ChatGPT.

Read the Full Blog Post for an In-Depth Exploration