Docs Menu

Docs HomeMongoDB Manual

Assign Weights to Text Search Results

On this page

  • About this Task
  • Before You Begin
  • Procedure
  • Results
  • Matches in content and about Fields
  • Matches in keywords and about Fields
  • Multiple Matches in a Single Document
  • Learn More

When MongoDB returns text search results, it assigns a score to each returned document. The score indicates the relevance of the document to a given search query. You can sort returned documents by score to have the most relevant documents appear first in the result set.

If you have a compound index with multiple text index keys, you can specify different weights for each indexed field. The weight of an indexed field indicates the significance of the field relative to the other indexed fields, with higher weights resulting in higher text search scores.

For example, you can emphasize search matches on a title field if you know users are likely to search for titles, or if title contains more relevant search terms compared to other document fields.

The default weight for indexed is 1 for the indexed fields. To adjust the weights for the indexed fields, include the weights option in the db.collection.createIndex() method, as seen in this example:

db.<collection>.createIndex(
{
<field1>: "text",
<field2>: "text",
...
},
{
weights: {
<field1>: <weight>,
<field2>: <weight>,
...
},
name: <indexName>
}
)

Important

If you change the weights in your index after it is created, MongoDB needs to reindex the collection. Reindexing can negatively impact performance, especially on large collections. For more information, see Index Builds on Populated Collections.

You have a blog collection that contains documents for individual blog posts. Each document contains:

  • The content of the post.

  • The topic that the post covers.

  • A list of keywords related to the post.

You want to create a text index so users can perform text searches on blog posts. Your application supports searches on content, topics, and keywords.

You want to prioritize matches on the content field over other document fields. Use index weights to assign greater importance to matches on content and sort query results so content matches appear first.

Create a blog collection with the following documents:

db.blog.insertMany( [
{
_id: 1,
content: "This morning I had a cup of coffee.",
about: "beverage",
keywords: [ "coffee" ]
},
{
_id: 2,
content: "Who likes chocolate ice cream for dessert?",
about: "food",
keywords: [ "poll" ]
},
{
_id: 3,
content: "My favorite flavors are strawberry and coffee",
about: "ice cream",
keywords: [ "food", "dessert" ]
}
] )

Create a text index with different weights for each indexed field:

db.blog.createIndex(
{
content: "text",
keywords: "text",
about: "text"
},
{
weights: {
content: 10,
keywords: 5
},
name: "BlogTextIndex"
}
)

The text index has the following fields and weights:

  • content has a weight of 10.

  • keywords has a weight of 5.

  • about has the default weight of 1.

These weights indicate the relative significance of the indexed fields to each other.

The following examples show how different weights for indexed fields affect result scores. Each example sorts results based on the textScore of each document. To access documents' textScore attributes, use the $meta operator.

The following query searches documents in the blog collection for the string ice cream:

db.blog.find(
{
$text: { $search: "ice cream" }
},
{
score: { $meta: "textScore" }
}
).sort( { score: { $meta: "textScore" } } )

Output:

[
{
_id: 2,
content: 'Who likes chocolate ice cream for dessert?',
about: 'food',
keywords: [ 'food', 'poll' ],
score: 12
},
{
_id: 3,
content: 'My favorite flavors are strawberry and coffee',
about: 'ice cream',
keywords: [ 'food', 'dessert' ],
score: 1.5
}
]

The search string ice cream matches:

  • The content field in the document with _id: 2.

  • The about field in the document with _id: 3.

A term match in the content field has 10 times the impact (10:1 weight) as a term match in the keywords field.

The following query searches documents in the blog collection for the string food:

db.blog.find(
{
$text: { $search: "food" }
},
{
score: { $meta: "textScore" }
}
).sort( { score: { $meta: "textScore" } } )

Output:

[
{
_id: 3,
content: 'My favorite flavors are strawberry and coffee',
about: 'ice cream',
keywords: [ 'food', 'dessert' ],
score: 5.5
},
{
_id: 2,
content: "Who likes chocolate ice cream for dessert?",
about: 'food',
keywords: [ 'poll' ],
score: 1.1
}
]

The search string food matches:

  • The keywords field in the document with _id: 3.

  • The about field in the document with _id: 2.

A term match in the keywords field has 5 times the impact (5:1 weight) as a term match in the about field.

The following query searches documents in the blog collection for the string coffee:

db.blog.find(
{
$text: { $search: "coffee" }
},
{
score: { $meta: "textScore" }
}
).sort( { score: { $meta: "textScore" } } )

Output:

[
{
_id: 1,
content: 'This morning I had a cup of coffee.',
about: 'beverage',
keywords: [ 'coffee' ],
score: 11.666666666666666
},
{
_id: 3,
content: 'My favorite cake flavors are strawberry and coffee',
about: 'ice cream',
keywords: [ 'food', 'dessert' ],
score: 6
}
]

The search string coffee matches:

  • The content and keywords fields in the document with _id: 1.

  • The content field in the document with _id: 3.

To calculate the score when a search string matches multiple fields, MongoDB multiplies the number of matches by the weight for the corresponding field and sums the results.

To learn more about text search in MongoDB, see:

Note

Atlas Search

For data hosted on MongoDB Atlas, Atlas Search provides more robust custom scoring than text indexes. To learn more, see the Atlas Search Scoring documentation.

←  Use Any Field to Specify Text Index LanguageLimit Number of Text Index Entries Scanned →