What is a fuzzy query in Elasticsearch and how does it handle typos?

Expected Interview Answer

Step-by-Step Explanation

1. Step 1

   Understand edit distance

   Levenshtein distance counts single-character insertions, deletions, substitutions, and transpositions between two terms.

2. Step 2

   Choose fuzziness

   Set fuzziness to an integer (0, 1, 2) or AUTO, which scales the allowed edits with the term's length.

3. Step 3

   Bound the cost

   Use prefix_length so the first characters must match exactly, and max_expansions to cap the number of variant terms.

4. Step 4

   Apply in a query

   Use the fuzzy query, or the match query with a fuzziness parameter, on an analyzed text field.

5. Step 5

   Tune for relevance

   Balance recall against precision — too much fuzziness returns noise, too little misses genuine typos.

Code Example

GET products/_search
{
  "query": {
    "fuzzy": {
      "title": {
        "value": "quikc",
        "fuzziness": "AUTO",
        "prefix_length": 1,
        "max_expansions": 50
      }
    }
  }
}

GET products/_search
{
  "query": {
    "match": {
      "title": {
        "query": "wireles keybard",
        "fuzziness": "AUTO"
      }
    }
  }
}

Follow-up Questions

• What does the AUTO fuzziness setting actually do for different term lengths?
• How do prefix_length and max_expansions affect performance?
• When would you choose match_phrase over a fuzzy query?
• How does a fuzzy query differ from a wildcard or regexp query?
• Can you combine fuzziness with a match query, and why would you?