Reverse Analyzer, Suffix or “Ends With” Search

By default, Elasticsearch matches from the start of words.
But what if you need to match from the end, for example, file extensions like .pdf or .zip?

Use a reverse token filter to flip the text, and then apply Edge N-Gram.

Example Mapping

PUT reverse_index
{
  "settings": {
    "analysis": {
      "filter": {
        "edge_reverse": {
          "type": "edge_ngram",
          "min_gram": 2,
          "max_gram": 10
        }
      },
      "analyzer": {
        "reverse_autocomplete": {
          "tokenizer": "standard",
          "filter": ["lowercase", "reverse", "edge_reverse", "reverse"]
        }
      }
    }
  },
  "mappings": {
    "properties": {
      "filename": {
        "type": "text",
        "analyzer": "reverse_autocomplete",
        "search_analyzer": "standard"
      }
    }
  }
}

Now, searching for “pdf” or “zip” will match “report.pdf” or “backup.zip”. Use for suffix-based search like extensions, domain endings, or last words.

Fuzzy Search, Handling Typos

Users often make spelling mistakes.
Fuzzy search helps Elasticsearch find results even with small typos or missing letters.

Concept

Fuzzy search uses Levenshtein edit distance, how many edits are needed to turn one word into another.

Examples:

• “iphon” → 1 edit away from “iphone”
• “googel” → 2 edits away from “google”

Example Query

GET products/_search
{
  "query": {
    "match": {
      "name": {
        "query": "iphon",
        "fuzziness": "AUTO"
      }
    }
  }
}

Matches “iPhone”, even though user typed “iphon”. You can also set fuzziness manually:

"fuzziness": 2

Performance Tip

Use fuzzy search only for short fields (titles, names),
because fuzzy queries are slower on long text fields.

Search-as-you-type — Built-In Autocomplete Field

From Elasticsearch 7.x onward, there’s an easier way:
the search_as_you_type field type — no need to define analyzers manually.

This field automatically creates small subfields (_2gram, _3gram, _index_prefix)
so it behaves like edge n-gram but optimized.

Example Mapping

PUT product_titles
{
  "mappings": {
    "properties": {
      "title": { "type": "search_as_you_type" }
    }
  }
}

Example Query

GET product_titles/_search
{
  "query": {
    "multi_match": {
      "query": "iph",
      "type": "bool_prefix",
      "fields": [
        "title",
        "title._2gram",
        "title._3gram",
        "title._index_prefix"
      ]
    }
  }
}

Matches “iPhone 15 Pro Max” as you type “iph”. Works with fuzziness: 1 for typo tolerance.

Combining Them for Best Experience

Feature	Goal	Ideal Analyzer / Query
Autocomplete (prefix)	Find as user types	Edge N-Gram / search_as_you_type
Suffix search	Match ending (e.g. .pdf)	Reverse Edge N-Gram
Typo tolerance	Handle spelling errors	Fuzzy search (fuzziness: AUTO)
Mid-word match	Search within words	N-Gram
Phrase match	Keep order	match_phrase or match_phrase_prefix

Example: All-In-One Setup

"mappings": {
  "properties": {
    "title": {
      "type": "text",
      "fields": {
        "autocomplete": { "type": "text", "analyzer": "autocomplete" },
        "reverse": { "type": "text", "analyzer": "reverse_autocomplete" },
        "keyword": { "type": "keyword" }
      }
    }
  }
}

Then use queries like:

{
  "bool": {
    "should": [
      { "match": { "title.autocomplete": "iphon" } },
      { "match": { "title.reverse": "pdf" } },
      { "match": { "title": { "query": "iphon", "fuzziness": 1 } } }
    ]
  }
}

Covers prefix, suffix, and typo in one unified search pipeline.