Character filters in Elasticsearch - Step-by-Step Execution

Concept Flow - Character filters

Input Text

↓

Apply Character Filters

↓

Modified Text

↓

Tokenizer

↓

Tokens for Analysis

Character filters take the input text and change characters before tokenizing, preparing text for analysis.

Execution Sample

Elasticsearch

{
  "settings": {
    "analysis": {
      "char_filter": {
        "my_filter": {
          "type": "mapping",
          "mappings": ["&=>and", "@=>at"]
        }
      },
      "analyzer": {
        "my_analyzer": {
          "type": "custom",
          "char_filter": ["my_filter"],
          "tokenizer": "whitespace"
        }
      }
    }
  }
}

This config replaces '&' with 'and' and '@' with 'at' before splitting text by spaces.

Execution Table

Step	Input Text	Character Filter Applied	Modified Text	Tokenizer Output
1	rock & roll @ night	Replace '&' with 'and'	rock and roll @ night	—
2	rock and roll @ night	Replace '@' with 'at'	rock and roll at night	—
3	rock and roll at night	No more filters	rock and roll at night	Tokens: ['rock', 'and', 'roll', 'at', 'night']
4	End	All filters applied	Final text ready for tokenizing	Tokenization complete

💡 All character filters applied; text is ready for tokenization.

Variable Tracker

Variable	Start	After Step 1	After Step 2	After Step 3	Final
input_text	rock & roll @ night	rock and roll @ night	rock and roll at night	rock and roll at night	rock and roll at night
modified_text	rock & roll @ night	rock and roll @ night	rock and roll at night	rock and roll at night	rock and roll at night
tokens	[]	[]	[]	['rock', 'and', 'roll', 'at', 'night']	['rock', 'and', 'roll', 'at', 'night']

Key Moments - 3 Insights

Why does the text change before tokenization?

Character filters modify the raw text first (see execution_table steps 1 and 2) so tokenization works on the cleaned text.

Are character filters applied after tokenization?

No, character filters run before tokenization (execution_table step 3), changing characters in the original text.

What happens if no character filters are defined?

The original text goes straight to tokenization without changes, so tokens reflect the raw input.

Visual Quiz - 3 Questions

Look at the execution_table at step 2, what is the modified text after replacing '@'?

A rock & roll at night

B rock and roll @ night

C rock and roll at night

D rock & roll @ night

💡 Hint

Check the 'Modified Text' column at step 2 in the execution_table.

At which step does tokenization produce the final tokens?

A Step 2

B Step 3

C Step 1

D Step 4

💡 Hint

Look at the 'Tokenizer Output' column in the execution_table.

If we remove the character filter replacing '&', what would be the token for '&'?

A '&'

B No token

C 'and'

D 'at'

💡 Hint

Refer to variable_tracker for 'tokens' after step 3 and consider what happens without the filter.

Concept Snapshot

Character filters modify input text before tokenizing.
They replace or remove characters.
Configured in analysis settings.
Run before tokenizer.
Help clean or normalize text.
Example: replace '&' with 'and'.

Full Transcript

Character filters in Elasticsearch change the input text before it is split into tokens. For example, they can replace symbols like '&' with words like 'and'. This happens before the tokenizer runs. The process starts with the original text, then each character filter applies its changes in order. After all filters run, the tokenizer splits the cleaned text into tokens. This helps search work better by normalizing text. If no character filters are used, the tokenizer works on the raw text. The example shows replacing '&' with 'and' and '@' with 'at', then splitting by spaces to get tokens.