Module 1: Let's Understand Search on Qdrant - Vector Search Engine

Module Overview

info@qdrant.tech (Andrey Vasnetsov) — Mon, 01 Jan 0001 00:00:00 +0000

Module 1

Let’s Understand Search

Traditional search retrieves documents by matching the words in a query to the words in an index. It’s fast and still useful for precise terms, but it can miss relevant results when people express the same intent differently. In this module, you’ll see how embeddings let search compare meaning, how similarity is measured, and why modern search often combines both approaches.

What Is Search?

info@qdrant.tech (Andrey Vasnetsov) — Mon, 01 Jan 0001 00:00:00 +0000

Module 1

What Is Search?

Search is the act of finding the right information out of everything you have, given a question. You type “car repair” into a box, and something has to decide which of your thousands of documents, products, or messages actually answers that.

Every search system, no matter how it’s built internally, does the same two things:

Retrieve: narrow a huge collection down to a shortlist of documents that might be relevant.
Rank: order that shortlist so the best answer ends up near the top.

At the heart of search is one question: what makes a result relevant? We’ll start with the simplest possible answer, watch it fail, and build up from there. No prior knowledge of vector search engines or indexing algorithms is assumed.

Why Keyword Search Struggles

info@qdrant.tech (Andrey Vasnetsov) — Mon, 01 Jan 0001 00:00:00 +0000

Module 1

Why Keyword Search Struggles

In its simplest form, keyword search retrieves documents by matching query terms against terms in the document. In other words, it matches the words themselves, not what they mean. It works when the query and document use the same vocabulary, but it can miss relevant results when the same intent is expressed differently.

How Traditional Search Improved

info@qdrant.tech (Andrey Vasnetsov) — Mon, 01 Jan 0001 00:00:00 +0000

Module 1

How Traditional Search Improved

Traditional search has evolved beyond exact word matching. Techniques such as stemming, typo tolerance, and relevance ranking make it faster and more forgiving. But they still rely on words: the system cannot tell that “car repair” and “automobile maintenance” mean the same thing unless that connection is explicitly defined.

That’s the gap semantic search closes. Instead of asking “Does this document contain the same words?” it asks “Does this document mean the same thing?” Nobody hand-codes the fact that “car” and “automobile” are related, the embedding model learns it from the text it was trained on, and sentences with related meaning end up as vectors that sit close together, even when they share no words.

How It Works: Embeddings

info@qdrant.tech (Andrey Vasnetsov) — Mon, 01 Jan 0001 00:00:00 +0000

Module 1

How It Works: Embeddings

What Is an Embedding?

An embedding is a vector: a list of numbers that captures meaning. Semantic search works by converting text into embeddings, text with similar meaning produces embeddings that sit close together in high-dimensional space, and text with different meaning produces embeddings that sit far apart. Each position in that list is a dimension; no single one maps to a human concept like “color,” meaning comes from all of them combined.

Comparing Meaning: Distance Metrics

info@qdrant.tech (Andrey Vasnetsov) — Mon, 01 Jan 0001 00:00:00 +0000

Module 1

Comparing Meaning: Distance Metrics

Once we have vectors, we need a way to measure how similar two of them are. Different metrics suit different situations.

Cosine Similarity

Cosine similarity measures the angle between two vectors. Vectors pointing in a similar direction score closer to 1. Vectors pointing in different directions score lower. It ignores vector length. For text embeddings, direction often carries more useful information about meaning than length does. This makes cosine similarity a common choice for semantic search.

Why Similarity Alone Is Not Enough

info@qdrant.tech (Andrey Vasnetsov) — Mon, 01 Jan 0001 00:00:00 +0000

Module 1

Why Similarity Alone Is Not Enough

What Is Search? and Why Keyword Search Struggles showed keyword search failing on synonyms, paraphrasing, polysemy, and word order. It’s tempting to read that as “semantic search replaces keyword search.” It doesn’t, each is strong exactly where the other is weak, as the next two cases show.

In Qdrant, each item you store is called a point. A point contains a vector and can also include a payload, which is metadata such as a timestamp or permission list. A collection is the group of points you search.

When a System Needs Both

info@qdrant.tech (Andrey Vasnetsov) — Mon, 01 Jan 0001 00:00:00 +0000

Module 1

When a System Needs Both

Those failures show why no single retrieval method works for every query. A search system may need semantic search to find related meaning, filters to match exact values such as a SKU, and keyword search when matching terms should affect the ranking.

Hybrid search specifically combines semantic retrieval with keyword retrieval. It uses dense vectors to compare meaning and sparse vectors to match and rank terms, often with an algorithm such as BM25.

Module 1: Let's Understand Search on Qdrant - Vector Search Engine

Module Overview

Let’s Understand Search

What Is Search?

What Is Search?

Why Keyword Search Struggles

Why Keyword Search Struggles

How Traditional Search Improved

How Traditional Search Improved

How It Works: Embeddings

How It Works: Embeddings

What Is an Embedding?

Comparing Meaning: Distance Metrics

Comparing Meaning: Distance Metrics

Cosine Similarity

Why Similarity Alone Is Not Enough

Why Similarity Alone Is Not Enough

When a System Needs Both

When a System Needs Both

Further Reading

Further Reading

What’s Next: Module 2