What Is Concept Search in eDiscovery?

Concept search, also called conceptual search or concept searching, sits in the analysis and review stages of the EDRM. It runs alongside keyword search, not instead of it, over the electronically stored information (ESI) a matter collects.

A keyword search matches literal strings of characters. It returns a document only when the exact term, or a stated variant, appears in the text. Concept search works one level up. It maps the relationships between words and retrieves documents about an idea, whatever vocabulary the author used.

The term is often used interchangeably with semantic search. Semantic search is the broader technology category. Concept search is the name eDiscovery practice uses for applying it to review and investigation.

How Does Concept Search Work?

Concept search engines build a mathematical model of how words relate to each other, then answer queries against that model. Most platforms follow the same basic sequence.

• The system indexes the collection and analyzes which terms appear together across documents.
• It builds a semantic model from those co-occurrence patterns, so related terms cluster around shared concepts.
• A query returns documents ranked by how closely they relate to the concept, not by literal term hits.

Some systems supplement this with a taxonomy or thesaurus that defines term relationships in advance. Statistical approaches, such as latent semantic indexing, learn the relationships from the dataset itself.

Concept clustering applies the same model in reverse. Instead of answering a query, it groups conceptually similar documents so reviewers can work through related material together. It also serves as a content filter during culling.

Concept Search vs Keyword Search

The two methods fail in different ways, which is why most matters use both.

Keyword search is precise but brittle. It misses synonyms, euphemisms, and code words, and it returns false hits when a term has several meanings. Research by Blair and Maron found that reviewers using keyword-based retrieval located only about 20 percent of the relevant documents, while believing they had found over 75 percent.

Concept search trades some precision for recall. It surfaces documents a keyword list could never anticipate, but its rankings still need human judgment to confirm relevance.

Keyword search remains the anchor of negotiated search protocols, since parties can agree on and test explicit terms. Concept search earns its place in investigation and early case assessment, where the vocabulary of a matter is still unknown.

Why Does Concept Search Matter?

People do not write to be found. Custodians use slang, shorthand, project code names, and euphemism, and no keyword list drafted in advance captures all of it. Every miss is a relevant document that never reaches review, which is a recall problem with defensibility consequences.

Concept search closes part of that gap. It also accelerates early case assessment, since clustering exposes the themes in a collection before anyone drafts search terms. And it feeds technology assisted review, which relies on the same underlying analysis of document meaning.

‍

Frequently Asked Questions

‍

What is concept search in eDiscovery?

Concept search is a method that retrieves documents based on meaning rather than exact keyword matches. In eDiscovery, it finds relevant ESI even when custodians used different words for the same idea.

What is the difference between concept search and keyword search?

Keyword search matches the literal terms in a query and misses documents that use other words for the same idea. Concept search retrieves documents by meaning, so synonyms and euphemisms still surface. Most matters use both together.

Is concept search the same as semantic search?

They describe the same underlying idea. Semantic search is the broader technology category, and concept search is the term eDiscovery practice uses for applying it to review and investigation.

Does concept search replace keyword search?

No. Keyword search remains the backbone of negotiated search protocols because explicit terms can be agreed and tested. Concept search complements it by catching the documents a keyword list cannot anticipate.