Analyzer components (tokenizer, filters) in Elasticsearch

Introduction

An analyzer breaks text into smaller pieces to help search engines understand it better. Tokenizers split text into words, and filters change or clean those words.

When you want to search text ignoring punctuation or special characters.

When you need to make searches case-insensitive so 'Apple' and 'apple' match.

When you want to remove common words like 'the' or 'and' to focus on important words.

When you want to break text into meaningful parts like words or numbers.

When you want to change words to their base form, like 'running' to 'run'.

Syntax

Elasticsearch

{
  "settings": {
    "analysis": {
      "analyzer": {
        "my_analyzer": {
          "type": "custom",
          "tokenizer": "standard",
          "filter": ["lowercase", "stop"]
        }
      },
      "filter": {
        "stop": {
          "type": "stop",
          "stopwords": ["the", "and"]
        }
      }
    }
  }
}

The tokenizer splits text into tokens (words).

The filter array applies changes to tokens, like making them lowercase or removing stop words.

Examples

This analyzer splits text by spaces and makes all words lowercase.

Elasticsearch

{
  "tokenizer": "whitespace",
  "filter": ["lowercase"]
}

This analyzer uses the standard tokenizer and removes common stop words like 'the' and 'and'.

Elasticsearch

{
  "tokenizer": "standard",
  "filter": ["lowercase", "stop"]
}

This analyzer treats the whole text as one token and makes it lowercase.

Elasticsearch

{
  "tokenizer": "keyword",
  "filter": ["lowercase"]
}

Sample Program

This example creates an index with a custom analyzer that splits text into words, makes them lowercase, and removes 'the' and 'and'. Then it analyzes a sentence to show the tokens.

Elasticsearch

PUT /my_index
{
  "settings": {
    "analysis": {
      "analyzer": {
        "my_custom_analyzer": {
          "type": "custom",
          "tokenizer": "standard",
          "filter": ["lowercase", "stop"]
        }
      },
      "filter": {
        "stop": {
          "type": "stop",
          "stopwords": ["the", "and"]
        }
      }
    }
  }
}

GET /my_index/_analyze
{
  "analyzer": "my_custom_analyzer",
  "text": "The quick brown fox jumps over the lazy dog and runs away"
}

Important Notes

Tokenizers break text into tokens, usually words or terms.

Filters can remove, change, or add tokens after tokenizing.

Stop filters remove common words that don't add meaning to searches.

Summary

Analyzers use tokenizers and filters to prepare text for searching.

Tokenizers split text into smaller pieces called tokens.

Filters modify tokens to improve search quality, like lowercasing or removing stop words.

Practice

1. What is the main role of a tokenizer in an Elasticsearch analyzer?

easy

A. To remove stop words from the text

B. To split text into smaller pieces called tokens

C. To convert tokens to lowercase

D. To index the tokens for searching

Solution

1. Step 1: Understand the tokenizer's function

   A tokenizer breaks the input text into smaller parts called tokens, which are the basic units for searching.

2. Step 2: Differentiate tokenizer from filters

   Filters modify tokens after tokenization, such as lowercasing or removing stop words, but tokenizers do the initial splitting.

3. Final Answer:

   To split text into smaller pieces called tokens -> Option B

4. Quick Check:

   Tokenizer = splits text [OK]

Hint: Tokenizer always splits text into tokens first [OK]

Common Mistakes:

• Confusing tokenizer with filters like lowercase or stop word removal
• Thinking tokenizer indexes tokens
• Assuming tokenizer modifies tokens after splitting

2. Which of the following is the correct syntax to define a standard tokenizer in an Elasticsearch analyzer settings?

easy

A. "tokenizer": standard

B. "tokenizer": ["standard"]

C. "tokenizer": { "type": "standard" }

D. "tokenizer": { "name": "standard" }

Solution

1. Step 1: Recall Elasticsearch tokenizer syntax

   Tokenizers are defined as objects with a type field specifying the tokenizer type.

2. Step 2: Compare options

   "tokenizer": { "type": "standard" } correctly uses an object with type key, while others use incorrect formats like strings or arrays.

3. Final Answer:

   "tokenizer": { "type": "standard" } -> Option C

4. Quick Check:

   Tokenizer syntax uses type key in object [OK]

Hint: Tokenizer must be an object with type key [OK]

Common Mistakes:

• Using string directly instead of object with type
• Using array syntax for tokenizer
• Using incorrect key names like name instead of type

3. Given this analyzer configuration snippet:

{
  "tokenizer": "whitespace",
  "filter": ["lowercase", "stop"]
}

What will be the output tokens for the input text: "The Quick Brown Fox"?

medium

A. ["quick", "brown", "fox"]

B. ["The", "Quick", "Brown", "Fox"]

C. ["quick", "brown"]

D. ["the", "quick", "brown", "fox"]

Solution

1. Step 1: Apply the whitespace tokenizer

   The whitespace tokenizer splits text by spaces, so tokens are ["The", "Quick", "Brown", "Fox"].

2. Step 2: Apply lowercase and stop filters

   The lowercase filter converts all tokens to lowercase: ["the", "quick", "brown", "fox"]. The stop filter removes common stop words like "the", leaving ["quick", "brown", "fox"].

3. Final Answer:

   ["quick", "brown", "fox"] -> Option A

4. Quick Check:

   Lowercase + stop filter removes "the" [OK]

Hint: Stop filter removes common words after lowercase [OK]

Common Mistakes:

• Ignoring stop word removal
• Not applying lowercase filter before stop filter
• Assuming tokenizer changes case

4. Identify the error in this analyzer filter configuration:

{
  "filter": [
    { "type": "lowercase" },
    { "type": "stop", "stopwords": "_english_" },
    { "type": "unknown_filter" }
  ]
}

medium

A. Filters cannot be defined as objects

B. The lowercase filter must be a string, not an object

C. The stop filter syntax is incorrect

D. The filter list contains an invalid filter type

Solution

1. Step 1: Review each filter type

   Lowercase and stop filters are valid and correctly defined. The stop filter uses a valid stopwords parameter.

2. Step 2: Identify invalid filter

   The filter with type "unknown_filter" is not a valid Elasticsearch filter type and will cause an error.

3. Final Answer:

   The filter list contains an invalid filter type -> Option D

4. Quick Check:

   Invalid filter type causes error [OK]

Hint: Check filter types are valid Elasticsearch filters [OK]

Common Mistakes:

• Assuming all filter types are valid
• Thinking lowercase filter must be string
• Ignoring unknown filter errors

5. You want to create a custom analyzer that splits text on whitespace, lowercases tokens, and removes English stop words. Which of the following analyzer settings is correct?

hard

A. { "tokenizer": { "type": "whitespace" }, "filter": ["lowercase", { "type": "stop", "stopwords": "_english_" }] }

B. { "tokenizer": "whitespace", "filter": ["lowercase", stop] }

C. { "tokenizer": "standard", "filter": ["lowercase", "stop"] }

D. { "tokenizer": { "type": "whitespace" }, "filter": ["stop", "lowercase"] }

Solution

1. Step 1: Confirm tokenizer type

   The tokenizer must be whitespace type, correctly defined as an object with type "whitespace".

2. Step 2: Confirm filter order and definition

   Filters should lowercase first, then remove stop words. { "tokenizer": { "type": "whitespace" }, "filter": ["lowercase", { "type": "stop", "stopwords": "_english_" }] } correctly defines lowercase as string and stop filter with explicit stopwords parameter.

3. Step 3: Check other options

   { "tokenizer": "whitespace", "filter": ["lowercase", "stop"] } uses tokenizer as string which is allowed but less explicit; { "tokenizer": "standard", "filter": ["lowercase", "stop"] } uses standard tokenizer which is incorrect; { "tokenizer": { "type": "whitespace" }, "filter": ["stop", "lowercase"] } reverses filter order which may affect results.

4. Final Answer:

   { "tokenizer": { "type": "whitespace" }, "filter": ["lowercase", { "type": "stop", "stopwords": "_english_" }] } -> Option A

5. Quick Check:

   Explicit tokenizer and correct filter order [OK]

Hint: Use object for tokenizer and correct filter order [OK]

Common Mistakes:

• Using wrong tokenizer type
• Reversing filter order
• Not specifying stopwords explicitly