Creating a Custom Analyzer

Sometimes, you need more control — maybe you want to:

• Remove accents (café → cafe)
• Add synonyms
• Use lowercase only
• Ignore specific words

Then you can define a custom analyzer inside index settings.

Example:

PUT my_index
{
  "settings": {
    "analysis": {
      "analyzer": {
        "my_custom_analyzer": {
          "type": "custom",
          "tokenizer": "standard",
          "filter": ["lowercase", "stop", "asciifolding"]
        }
      }
    }
  },
  "mappings": {
    "properties": {
      "title": { "type": "text", "analyzer": "my_custom_analyzer" }
    }
  }
}

How It Works:

1. Tokenizer: breaks text into words.
2. Lowercase filter: converts all to lowercase.
3. Stop filter: removes common words (“the”, “a”).
4. Asciifolding: removes accents (café → cafe).

Now when you index:

POST my_index/_doc
{ "title": "Café in Jakarta" }

and search for "cafe", it will match perfectly — even without the accent.

Adding Synonyms

You can also teach Elasticsearch that some words mean the same thing.
For example: “car”, “auto”, “vehicle”.

PUT synonym_index
{
  "settings": {
    "analysis": {
      "filter": {
        "synonym_filter": {
          "type": "synonym",
          "synonyms": [
            "car, automobile, vehicle"
          ]
        }
      },
      "analyzer": {
        "my_synonym_analyzer": {
          "tokenizer": "standard",
          "filter": ["lowercase", "synonym_filter"]
        }
      }
    }
  }
}

Now searching "automobile" will also find "car" and "vehicle".

Synonyms make search more human, users don’t always type the same word you used.

Comparison: Standard vs Custom Analyzer

Feature	Standard Analyzer	Custom Analyzer
Predefined	Yes	No
Easy to use	Yes	Needs config
Language support	Basic	Flexible
Synonyms	No	Yes
Filters	Limited	Any combination
Best for	General text	Domain-specific search

Analyzer Best Practices

• Always lowercase your text: case differences don’t matter in most searches.
• Use language analyzers when your data has meaningful word forms.
• Add stopwords carefully: removing too many words may break phrases.
• Test with _analyze before indexing large data.
• Use search_analyzer if you want a different analyzer at query time.

Example:

"analyzer": "my_custom_analyzer",
"search_analyzer": "standard"

Indexes text using custom rules but searches more broadly.

How Analyzers Affect Search and Autocomplete

Even though analyzers seem low-level, they’re the foundation of autocomplete and fuzzy matching.

For example:

• An analyzer with lowercase + edge n-gram: enables “search as you type”
• A language analyzer: improves semantic search
• A custom analyzer with synonyms: improves query recall

So, understanding analyzers = understanding how Elasticsearch finds what users mean, not just what they type.

Conclusion

Analyzers are the brain of Elasticsearch text search. They define how your text is split, normalized, and matched, and with a bit of customization, they can make your search engine smarter than ever.

“If mapping is the skeleton, analyzer is the brain.”

Start with the built-in ones (standard, english, indonesian),
and once you understand how they tokenize text,
experiment with your own analyzer to match your domain’s language and behavior.