Token filters (lowercase, stemmer, synonym) in Elasticsearch

Introduction

Token filters change words in text to help search work better. They make words lowercase, find word roots, or add similar words.

You want to find words no matter if they are uppercase or lowercase.

You want to match different forms of a word like 'run', 'running', 'runs'.

You want to include similar words like 'car' and 'automobile' in search results.

You want to improve search by handling synonyms automatically.

You want to make search results more flexible and user-friendly.

Syntax

Elasticsearch

PUT /my_index
{
  "settings": {
    "analysis": {
      "filter": {
        "my_lowercase": {
          "type": "lowercase"
        },
        "my_stemmer": {
          "type": "stemmer",
          "language": "english"
        },
        "my_synonym": {
          "type": "synonym",
          "synonyms": [
            "car, automobile",
            "quick, fast"
          ]
        }
      },
      "analyzer": {
        "my_custom_analyzer": {
          "tokenizer": "standard",
          "filter": ["my_lowercase", "my_stemmer", "my_synonym"]
        }
      }
    }
  }
}

The lowercase filter makes all words small letters.

The stemmer filter cuts words to their root form.

The synonym filter adds words that mean the same.

Examples

This filter changes all tokens to lowercase.

Elasticsearch

"filter": { "lowercase_filter": { "type": "lowercase" } }

This filter reduces English words to their root form.

Elasticsearch

"filter": { "english_stemmer": { "type": "stemmer", "language": "english" } }

This filter treats listed words as synonyms during search.

Elasticsearch

"filter": { "synonym_filter": { "type": "synonym", "synonyms": ["fast, quick", "car, automobile"] } }

Sample Program

This example creates an index with lowercase, stemmer, and synonym filters. Then it analyzes the text 'Cars are running fast' to show how tokens change.

Elasticsearch

PUT /example_index
{
  "settings": {
    "analysis": {
      "filter": {
        "lowercase_filter": {
          "type": "lowercase"
        },
        "english_stemmer": {
          "type": "stemmer",
          "language": "english"
        },
        "synonym_filter": {
          "type": "synonym",
          "synonyms": ["car, automobile", "fast, quick"]
        }
      },
      "analyzer": {
        "custom_analyzer": {
          "tokenizer": "standard",
          "filter": ["lowercase_filter", "english_stemmer", "synonym_filter"]
        }
      }
    }
  }
}

GET /example_index/_analyze
{
  "analyzer": "custom_analyzer",
  "text": "Cars are running fast"
}

Important Notes

Order of filters matters: lowercase should come before stemmer and synonym.

Synonym filter can use external files for large synonym lists.

Stemming may reduce words too much, so test your filters carefully.

Summary

Token filters change words to improve search matching.

Lowercase filter makes all words small letters.

Stemmer filter finds root forms of words.

Synonym filter adds similar words to search.

Practice

1.

What does the lowercase token filter do in Elasticsearch?

easy

A. Converts all tokens to lowercase letters

B. Removes stop words from the tokens

C. Finds the root form of tokens

D. Adds synonyms to tokens

Solution

1. Step 1: Understand the lowercase filter purpose

   The lowercase filter changes all tokens to lowercase to make searches case-insensitive.

2. Step 2: Compare with other filters

   Stemmer finds root words, synonym adds similar words, stop word filter removes common words. Only lowercase converts to lowercase.

3. Final Answer:

   Converts all tokens to lowercase letters -> Option A

4. Quick Check:

   Lowercase filter = converts tokens to lowercase [OK]

Hint: Lowercase filter always makes words small letters [OK]

Common Mistakes:

• Confusing lowercase with stemmer
• Thinking lowercase adds synonyms
• Mixing lowercase with stop word removal

2.

Which of the following is the correct syntax to define a synonym token filter in Elasticsearch?

{
  "filter": {
    "my_synonym": {
      "type": "synonym",
      "synonyms": ["quick,fast"]
    }
  }
}

easy

A. { "type": "stemmer", "synonyms": ["quick,fast"] }

B. { "type": "stop", "synonyms": ["quick,fast"] }

C. { "type": "synonym", "synonyms": ["quick,fast"] }

D. { "type": "lowercase", "synonyms": ["quick,fast"] }

Solution

1. Step 1: Identify the token filter type for synonyms

   The synonym filter must have "type": "synonym" to work correctly.

2. Step 2: Check the synonyms property

   The synonyms list is correctly provided as an array of strings with comma-separated words.

3. Final Answer:

   { "type": "synonym", "synonyms": ["quick,fast"] } -> Option C

4. Quick Check:

   Synonym filter syntax = type synonym with synonyms list [OK]

Hint: Synonym filter always uses type 'synonym' with synonyms array [OK]

Common Mistakes:

• Using wrong type like stemmer or lowercase
• Not providing synonyms as an array
• Misspelling 'synonym' in type

3.

Given this analyzer configuration in Elasticsearch:

{
  "analyzer": {
    "my_analyzer": {
      "tokenizer": "standard",
      "filter": ["lowercase", "porter_stem"]
    }
  }
}

What will be the output tokens for the input text "Running Runs"?

medium

A. ["run", "runs"]

B. ["running", "runs"]

C. ["Running", "Runs"]

D. ["run", "run"]

Solution

1. Step 1: Apply lowercase filter

   Input "Running Runs" becomes "running runs" after lowercase filter.

2. Step 2: Apply porter_stem filter

   Porter stemmer reduces "running" and "runs" to root "run" for both tokens.

3. Final Answer:

   ["run", "run"] -> Option D

4. Quick Check:

   Lowercase + stemmer = root forms in lowercase [OK]

Hint: Lowercase then stemmer gives root lowercase tokens [OK]

Common Mistakes:

• Ignoring stemming effect
• Not applying lowercase first
• Expecting original case tokens

4.

Identify the error in this token filter definition:

{
  "filter": {
    "my_stemmer": {
      "type": "stemmer",
      "language": "englishy"
    }
  }
}

medium

A. "type" should be "stemmer_override"

B. "language" value is misspelled; should be "english"

C. "filter" key must be inside "analyzer"

D. Missing "synonyms" array

Solution

1. Step 1: Check the "language" parameter spelling

   The language "englishy" is incorrect; the correct value is "english".

2. Step 2: Verify the stemmer type and parameters

   Type "stemmer" is correct; no synonyms needed here; filter can be defined outside analyzer.

3. Final Answer:

   "language" value is misspelled; should be "english" -> Option B

4. Quick Check:

   Stemmer language spelling must be exact [OK]

Hint: Check spelling of language names in stemmer filters [OK]

Common Mistakes:

• Using invalid language names
• Confusing stemmer with synonym filter
• Expecting synonyms in stemmer filter

5.

You want to create an analyzer that makes searches case-insensitive, finds root forms of words, and also matches synonyms like "fast" for "quick". Which token filters should you include in your analyzer?

hard

A. ["lowercase", "stemmer", "synonym"]

B. ["stop", "lowercase", "synonym"]

C. ["lowercase", "synonym"]

D. ["stemmer", "stop"]

Solution

1. Step 1: Identify filters for case-insensitivity and root forms

   Lowercase filter makes tokens lowercase; stemmer finds root forms.

2. Step 2: Add synonym filter for matching similar words

   Synonym filter adds synonyms like "fast" for "quick" to improve matching.

3. Final Answer:

   ["lowercase", "stemmer", "synonym"] -> Option A

4. Quick Check:

   Lowercase + stemmer + synonym = case-insensitive, root, and synonym matching [OK]

Hint: Combine lowercase, stemmer, and synonym filters for full matching [OK]

Common Mistakes:

• Omitting stemmer or synonym filters
• Using stop filter instead of stemmer
• Not including lowercase for case-insensitivity