Docs Menu

Docs Home → Launch & Manage MongoDB → MongoDB Atlas

Tokenizers

On this page

  • edgeGram
  • Attributes
  • Example
  • keyword
  • Attributes
  • Example
  • nGram
  • Attributes
  • Example
  • regexCaptureGroup
  • Attributes
  • Example
  • regexSplit
  • Attributes
  • Example
  • standard
  • Attributes
  • Example
  • uaxUrlEmail
  • Attributes
  • Example
  • whitespace
  • Attributes
  • Example

A custom analyzer's tokenizer determines how Atlas Search splits up text into discrete chunks for indexing. Tokenizers require a type field, and some take additional options as well.

"tokenizer": {
  "type": "<tokenizer-type>",
  "<additional-option>": "<value>"
}

The following sample index definitions and queries use the sample collection named minutes. If you add the minutes collection to a database in your Atlas cluster, you can create the following sample indexes from the Visual Editor or JSON Editor in the Atlas UI and run the sample queries against this collection. To create these indexes, after you select your preferred configuration method in the Atlas UI, select the database and collection, and refine your index to add custom analyzers that use tokenizers.

Note

When you add a custom analyzer using the Visual Editor in the Atlas UI, the Atlas UI displays the following details about the analyzer in the Custom Analyzers section.

Name
Label that identifies the custom analyzer.
Used In
Fields that use the custom analyzer. Value is None if custom analyzer isn't used to analyze any fields.
Character Filters
Atlas Search character filters configured in the custom analyzer.
Tokenizer
Atlas Search tokenizer configured in the custom analyzer.
Token Filters
Atlas Search token filters configured in the custom analyzer.
Actions

Clickable icons that indicate the actions that you can perform on the custom analyzer.

  • Click to edit the custom analyzer.

  • Click to delete the custom analyzer.

edgeGram

The edgeGram tokenizer tokenizes input from the left side, or "edge", of a text input into n-grams of given sizes. You can't use a custom analyzer with edgeGram tokenizer in the analyzer field for synonym or autocomplete field mapping definitions.

Attributes

It has the following attributes:

Note

The edgeGram tokenizer yields multiple output tokens per word and across words in input text, producing token graphs.

Because autocomplete field type mapping definitions and analyzers with synonym mappings only work when used with non-graph-producing tokenizers, you can't use a custom analyzer with edgeGram tokenizer in the analyzer field for autocomplete field type mapping definitions or analyzers with synonym mappings.

Name
Type
Required?
Description
type
string
yes
Human-readable label that identifies this tokenizer type. Value must be edgeGram.
minGram
integer
yes
Number of characters to include in the shortest token created.
maxGram
integer
yes
Number of characters to include in the longest token created.

Example

The following index definition indexes the message field in the minutes collection using a custom analyzer named edgegramExample. It uses the edgeGram tokenizer to create tokens (searchable terms) between 2 and 7 characters long starting from the first character on the left side of words in the message field.

The following query searches the message field in the minutes collection for text that begin with tr.

1db.minutes.aggregate([
2  {
3    "$search": {
4      "text": {
5        "query": "tr",
6        "path": "message"
7      }
8    }
9  },
10  {
11    "$project": {
12      "_id": 1,
13      "message": 1
14    }
15  }
16])
{ _id: 1, message: 'try to siGn-In' },
{ _id: 3, message: 'try to sign-in' }

Atlas Search returns documents with _id: 1 and _id: 3 in the results because Atlas Search created a token with the value tr using the edgeGram tokenizer for the documents, which matches the search term. If you index the message field using the standard tokenizer, Atlas Search would not return any results for the search term tr.

The following table shows the tokens that the edgeGram tokenizer and by comparison, the standard tokenizer, create for the documents in the results:

Tokenizer
Token Outputs
standard
try, to, sign, in
edgeGram
tr, try, try{SPACE}, try t, try to, try to{SPACE}

keyword

The keyword tokenizer tokenizes the entire input as a single token. Atlas Search doesn't index string fields that exceed 32766 characters using the keyword tokenizer.

Attributes

It has the following attributes:

Name
Type
Required?
Description
type
string
yes
Human-readable label that identifies this tokenizer type. Value must be keyword.

Example

The following index definition indexes the message field in the minutes collection using a custom analyzer named keywordExample. It uses the keyword tokenizer to create a token (searchable terms) on the entire field as a single term.

The following query searches the message field in the minutes collection for the phrase try to sign-in.

1db.minutes.aggregate([
2  {
3    "$search": {
4      "text": {
5        "query": "try to sign-in",
6        "path": "message"
7      }
8    }
9  },
10  {
11    "$project": {
12      "_id": 1,
13      "message": 1
14    }
15  }
16])
{ _id: 3, message: 'try to sign-in' }

Atlas Search returns the document with _id: 3 in the results because Atlas Search created a token with the value try to sign-in using the keyword tokenizer for the documents, which matches the search term. If you index the message field using the standard tokenizer, Atlas Search returns documents with _id: 1, _id: 2 and _id: 3 for the search term try to sign-in because each document contains some of the tokens the standard tokenizer creates.

The following table shows the tokens that the keyword tokenizer and by comparison, the standard tokenizer, create for the document with _id: 3:

Tokenizer
Token Outputs
standard
try, to, sign, in
keyword
try to sign-in

nGram

The nGram tokenizer tokenizes into text chunks, or "n-grams", of given sizes. You can't use a custom analyzer with nGram tokenizer in the analyzer field for synonym or autocomplete field mapping definitions.

Attributes

It has the following attributes:

Name
Type
Required?
Description
type
string
yes
Human-readable label that identifies this tokenizer type. Value must be nGram.
minGram
integer
yes
Number of characters to include in the shortest token created.
maxGram
integer
yes
Number of characters to include in the longest token created.

Example

The following index definition indexes the title field in the minutes collection using a custom analyzer named ngramExample. It uses the nGram tokenizer to create tokens (searchable terms) between 4 and 6 characters long in the title field.

The following query searches the title field in the minutes collection for the term week.

1db.minutes.aggregate([
2  {
3    "$search": {
4      "text": {
5        "query": "week",
6        "path": "title"
7      }
8    }
9  },
10  {
11    "$project": {
12      "_id": 1,
13      "title": 1
14    }
15  }
16])
{ _id: 1, title: "The team's weekly meeting" }

Atlas Search returns the document with _id: 1 in the results because Atlas Search created a token with the value week using the nGram tokenizer for the documents, which matches the search term. If you index the title field using the standard or edgeGram tokenizer, Atlas Search would not return any results for the search term week.

The following table shows the tokens that the nGram tokenizer and by comparison, the standard and edgeGram tokenizer create for the document with _id: 1:

Tokenizer
Token Outputs
standard
The, team's, weekly, meeting
edgeGram
The{SPACE}, The t, The te
nGram
The{SPACE}, The t, The te, he t, ... , week, weekl, weekly, eekl, ..., eetin, eeting, etin, eting, ting

regexCaptureGroup

The regexCaptureGroup tokenizer matches a regular expression pattern to extract tokens.

Attributes

It has the following attributes:

Name
Type
Required?
Description
type
string
yes
Human-readable label that identifies this tokenizer type. Value must be regexCaptureGroup.
pattern
string
yes
Regular expression to match against.
group
integer
yes
Index of the character group within the matching expression to extract into tokens. Use 0 to extract all character groups.

Example

The following index definition indexes the page_updated_by.phone field in the minutes collection using a custom analyzer named phoneNumberExtractor. It uses the following:

  • mappings character filter to remove parenthesis around the first three digits and replace all spaces and periods with dashes

  • regexCaptureGroup tokenizer to create a single token from the first US-formatted phone number present in the text input

The following query searches the page_updated_by.phone field in the minutes collection for the phone number 123-456-9870.

1db.minutes.aggregate([
2  {
3    "$search": {
4      "text": {
5        "query": "123-456-9870",
6        "path": "page_updated_by.phone"
7      }
8    }
9  },
10  {
11    "$project": {
12      "_id": 1,
13      "page_updated_by.phone": 1
14    }
15  }
16])
{ _id: 3, page_updated_by: { phone: '(123).456.9870' }

Atlas Search returns the document with _id: 3 in the results because Atlas Search created a token with the value 123-456-7890 using the regexCaptureGroup tokenizer for the documents, which matches the search term. If you index the page_updated_by.phone field using the standard tokenizer, Atlas Search returns all of the documents for the search term 123-456-7890.

The following table shows the tokens that the regexCaptureGroup tokenizer and by comparison, the standard tokenizer, create for the document with _id: 3:

Tokenizer
Token Outputs
standard
123, 456.9870
regexCaptureGroup
123-456-9870

regexSplit

The regexSplit tokenizer splits tokens with a regular-expression based delimiter.

Attributes

It has the following attributes:

Name
Type
Required?
Description
type
string
yes
Human-readable label that identifies this tokenizer type. Value must be regexSplit.
pattern
string
yes
Regular expression to match against.

Example

The following index definition indexes the page_updated_by.phone field in the minutes collection using a custom analyzer named dashDotSpaceSplitter. It uses the regexSplit tokenizer to create tokens (searchable terms) from one or more hyphens, periods and spaces on the page_updated_by.phone field.

The following query searches the page_updated_by.phone field in the minutes collection for the digits 9870.

1db.minutes.aggregate([
2  {
3    "$search": {
4      "text": {
5        "query": "9870",
6        "path": "page_updated_by.phone"
7      }
8    }
9  },
10  {
11    "$project": {
12      "_id": 1,
13      "page_updated_by.phone": 1
14    }
15  }
16])
{ _id: 3, page_updated_by: { phone: '(123).456.9870' }

Atlas Search returns the document with _id: 3 in the results because Atlas Search created a token with the value 9870 using the regexSplit tokenizer for the documents, which matches the search term. If you index the page_updated_by.phone field using the standard tokenizer, Atlas Search would not return any results for the search term 9870.

The following table shows the tokens that the regexCaptureGroup tokenizer and by comparison, the standard tokenizer, create for the document with _id: 3:

Tokenizer
Token Outputs
standard
123, 456.9870
regexSplit
(123), 456, 9870

standard

The standard tokenizer tokenizes based on word break rules from the Unicode Text Segmentation algorithm.

Attributes

It has the following attributes:

Name
Type
Required?
Description
type
string
yes
Human-readable label that identifies this tokenizer type. Value must be standard.
maxTokenLength
integer
no

Maximum length for a single token. Tokens greater than this length are split at maxTokenLength into multiple tokens.

Default: 255

Example

The following index definition indexes the message field in the minutes collection using a custom analyzer named standardExample. It uses the standard tokenizer and the stopword token filter.

The following query searches the message field in the minutes collection for the term signature.

1db.minutes.aggregate([
2  {
3    "$search": {
4      "text": {
5        "query": "signature",
6        "path": "message"
7      }
8    }
9  },
10  {
11    "$project": {
12      "_id": 1,
13      "message": 1
14    }
15  }
16])
{ _id: 4, message: 'write down your signature or phone №' }

Atlas Search returns the document with _id: 4 because Atlas Search created a token with the value signature using the standard tokenizer for the documents, which matches the search term. If you index the message field using the keyword tokenizer, Atlas Search would not return any results for the search term signature.

The following table shows the tokens that the standard tokenizer and by comparison, the keyword analyzer, create for the document with _id: 4:

Tokenizer
Token Outputs
standard
write, down, your, signature, or, phone
keyword
write down your signature or phone №

uaxUrlEmail

The uaxUrlEmail tokenizer tokenizes URLs and email addresses. Although uaxUrlEmail tokenizer tokenizes based on word break rules from the Unicode Text Segmentation algorithm, we recommend using uaxUrlEmail tokenizer only when the indexed field value includes URLs and email addresses. For fields that don't include URLs or email addresses, use the standard tokenizer to create tokens based on word break rules.

Attributes

It has the following attributes:

Name
Type
Required?
Description
type
string
yes
Human-readable label that identifies this tokenizer type. Value must be uaxUrlEmail.
maxTokenLength
int
no

Maximum number of characters in one token.

Default: 255

Example

whitespace

The whitespace tokenizer tokenizes based on occurrences of whitespace between words.

Attributes

It has the following attributes:

Name
Type
Required?
Description
type
string
yes
Human-readable label that identifies this tokenizer type. Value must be whitespace.
maxTokenLength
integer
no

Maximum length for a single token. Tokens greater than this length are split at maxTokenLength into multiple tokens.

Default: 255

Example

The following index definition indexes the message field in the minutes collection using a custom analyzer named whitespaceExample. It uses the whitespace tokenizer to create tokens (searchable terms) from any whitespaces in the message field.

The following query searches the message field in the minutes collection for the term SIGN-IN.

1db.minutes.aggregate([
2  {
3    "$search": {
4      "text": {
5        "query": "SIGN-IN",
6        "path": "message"
7      }
8    }
9  },
10  {
11    "$project": {
12      "_id": 1,
13      "message": 1
14    }
15  }
16])
{ _id: 2, message: 'do not forget to SIGN-IN' }

Atlas Search returns the document with _id: 2 in the results because Atlas Search created a token with the value SIGN-IN using the whitespace tokenizer for the documents, which matches the search term. If you index the message field using the standard tokenizer, Atlas Search returns documents with _id: 1, _id: 2 and _id: 3 for the search term SIGN-IN.

The following table shows the tokens that the whitespace tokenizer and by comparison, the standard tokenizer, create for the document with _id: 2:

Tokenizer
Token Outputs
standard
do, not, forget, to, sign, in
whitespace
do, not, forget, to, SIGN-IN
← Character Filters
Token Filters →

On this page