Modern search systems don’t just find exact matches, they understand partial words, typos, and phrases as you type. Elasticsearch makes this possible with a mix of analyzers and special queries like N-Gram, Reverse, Fuzzy, and Search-as-you-type.

In this article, we’ll learn how these techniques work, when to use each, and how to combine them for a fast and smart search experience.

Why “Advanced” Text Search?

In real-world apps, users rarely type exact words.
They type fast, make mistakes, or expect results before finishing the query.

For example:

• “iph”, should match “iPhone 15 Pro”
• “iphon”, should still match “iPhone” (even with typo)
• “.pdf”, should find “report.pdf”

To make that work, Elasticsearch uses special analyzers and queries under the hood.

N-Gram Analyzer — Matching Inside Words

The N-Gram analyzer splits text into small overlapping pieces called n-grams.
This allows search to match any part of a word, not just the beginning.

Example

Text: "search"
With min_gram: 3, max_gram: 5, tokens become:

["sea", "ear", "arc", "rch"]

So a user typing “arc” still finds “search”.

Example Mapping

PUT ngram_index
{
  "settings": {
    "analysis": {
      "tokenizer": {
        "my_ngram": {
          "type": "ngram",
          "min_gram": 3,
          "max_gram": 5
        }
      },
      "analyzer": {
        "my_ngram_analyzer": {
          "tokenizer": "my_ngram",
          "filter": ["lowercase"]
        }
      }
    }
  },
  "mappings": {
    "properties": {
      "name": { "type": "text", "analyzer": "my_ngram_analyzer" }
    }
  }
}

Now when you search:

GET ngram_index/_search
{
  "query": { "match": { "name": "arc" } }
}

It matches “search”, “arctic”, “arcade”, etc.

Trade-off

N-Gram creates many tokens, bigger index size.
Use it only for short fields like names or titles, not large documents.

Edge N-Gram, Perfect for Autocomplete

Edge N-Gram works like N-Gram, but it only creates tokens from the start of the word.
This makes it ideal for prefix matching (autocomplete).

Text: "search"
Edge N-Gram tokens (min_gram=2, max_gram=5):

["se", "sea", "sear", "searc"]

So when a user types “sear”, it matches “search”.

Example Mapping

PUT edge_index
{
  "settings": {
    "analysis": {
      "tokenizer": {
        "edge_tokenizer": {
          "type": "edge_ngram",
          "min_gram": 2,
          "max_gram": 10
        }
      },
      "analyzer": {
        "autocomplete": {
          "tokenizer": "edge_tokenizer",
          "filter": ["lowercase"]
        }
      }
    }
  },
  "mappings": {
    "properties": {
      "title": {
        "type": "text",
        "analyzer": "autocomplete",
        "search_analyzer": "standard"
      }
    }
  }
}

Now, as you type:

• “se”: matches “search”, “service”, “secure”
• “serv”: matches “server”, “service”

Great for instant autocomplete search bars.