Fuzzy matching in Elasticsearch - Step-by-Step Execution

Concept Flow - Fuzzy matching

Input Query

↓

Apply Fuzziness

↓

Search Index for Similar Terms

↓

Calculate Similarity Score

↓

Return Matches Above Threshold

↓

Display Results

The search query is processed with fuzziness to find terms similar to the input, then results with high similarity scores are returned.

Execution Sample

Elasticsearch

{
  "query": {
    "fuzzy": {
      "name": {
        "value": "roam",
        "fuzziness": "AUTO"
      }
    }
  }
}

This query searches for documents where the 'name' field matches terms similar to 'roam' using automatic fuzziness.

Execution Table

Step	Action	Input	Fuzziness Applied	Similarity Score	Result
1	Receive query	name: 'roam'	AUTO	-	Start search
2	Generate variants	roam	AUTO	-	roam, room, foam, roam...
3	Search index	variants	AUTO	Calculated per term	Find matching documents
4	Calculate similarity	each variant vs index terms	AUTO	0.8, 0.9, 0.7, ...	Score each match
5	Filter results	scores	AUTO	>= threshold	Keep matches with high score
6	Return results	filtered matches	AUTO	-	Documents with similar terms to 'roam'
7	End	-	-	-	Search complete

💡 Search ends after returning documents with similarity scores above threshold.

Variable Tracker

Variable	Start	After Step 2	After Step 3	After Step 4	Final
query	{name: 'roam'}	{name: 'roam'}	{variants generated}	{scores calculated}	{filtered matches}
variants	none	[roam, room, foam]	[roam, room, foam]	[roam, room, foam]	[roam, room]
similarity_scores	none	none	none	[0.9, 0.8, 0.7]	[0.9, 0.8]
results	none	none	none	none	[docs matching 'roam' and 'room']

Key Moments - 3 Insights

Why does the query find 'room' when searching for 'roam'?

Because fuzziness allows small differences, the execution_table row 2 shows variants like 'room' generated from 'roam'.

What does the similarity score represent?

It measures how close a variant term is to the original query term, as shown in execution_table row 4.

Why are some variants filtered out?

Only variants with similarity scores above a threshold are kept, as explained in execution_table row 5.

Visual Quiz - 3 Questions

Look at the execution_table, what is the similarity score for the variant 'room'?

A 0.8

B 0.9

C 0.7

D 1.0

💡 Hint

Check execution_table row 4 under Similarity Score column.

At which step are variants like 'foam' generated from 'roam'?

A Step 2

B Step 4

C Step 1

D Step 6

💡 Hint

Look at execution_table row 2 describing variant generation.

If fuzziness was set to 0, how would the variants change?

A More variants generated

B Variants with large differences generated

C No variants generated, exact match only

D Variants ignored, all results returned

💡 Hint

Fuzziness controls allowed differences; 0 means exact match only, see concept_flow step 'Apply Fuzziness'.

Concept Snapshot

Fuzzy matching in Elasticsearch:
- Uses 'fuzzy' query with 'value' and 'fuzziness' params
- Finds terms similar to input (typos, close spellings)
- Generates variants automatically with 'AUTO' fuzziness
- Scores similarity and filters results
- Useful for typo-tolerant search

Full Transcript

Fuzzy matching in Elasticsearch works by taking the input query term and generating similar variants based on allowed differences called fuzziness. The query then searches the index for these variants, calculates similarity scores for each match, and returns documents with scores above a threshold. For example, searching for 'roam' with fuzziness 'AUTO' generates variants like 'room' and 'foam'. Each variant is scored for similarity, and only close matches are returned. This helps find results even if the search term has typos or small differences. The process starts with receiving the query, generating variants, searching the index, scoring matches, filtering results, and finally returning the matched documents.