Language Analyzers

On this page

Examples

Built-In Language Analyzer Example
Custom Language Analyzer Example

Use language-specific analyzers to create indexes tailored to a particular language. Each language analyzer has built-in stop words and word divisions based on that language's usage patterns.

Atlas Search offers the following language analyzers:

`lucene.arabic`	`lucene.armenian`	`lucene.basque`	`lucene.bengali`
`lucene.brazilian`	`lucene.bulgarian`	`lucene.catalan`	`lucene.chinese`
`lucene.cjk` ¹	`lucene.czech`	`lucene.danish`	`lucene.dutch`
`lucene.english`	`lucene.finnish`	`lucene.french`	`lucene.galician`
`lucene.german`	`lucene.greek`	`lucene.hindi`	`lucene.hungarian`
`lucene.indonesian`	`lucene.irish`	`lucene.italian`	`lucene.japanese`
`lucene.korean`	`lucene.kuromoji` ²	`lucene.latvian`	`lucene.lithuanian`
`lucene.morfologik` ³	`lucene.nori` ⁴	`lucene.norwegian`	`lucene.persian`
`lucene.polish`	`lucene.portuguese`	`lucene.romanian`	`lucene.russian`
`lucene.smartcn` ⁵	`lucene.sorani`	`lucene.spanish`	`lucene.swedish`
`lucene.thai`	`lucene.turkish`	`lucene.ukrainian`

¹ cjk is a generic Chinese, Japanese, and Korean analyzer

² kuromoji is a Japanese analyzer

³ morfologik is a Polish analyzer

⁴ nori is a Korean analyzer

⁵ smartcn is a Chinese analyzer

Examples

Consider a collection named cars with the following documents:

{
  "_id": 1,
  "subject": {
    "en": "It is better to equip our cars to understand the causes of the accident.",
    "fr": "Mieux équiper nos voitures pour comprendre les causes d'un accident.",
    "he": "עדיף לצייד את המכוניות שלנו כדי להבין את הגורמים לתאונה."
  }
}

{
  "_id": 2,
  "subject": {
    "en": "The best time to do this is immediately after you've filled up with fuel",
    "fr": "Le meilleur moment pour le faire c'est immédiatement après que vous aurez fait le plein de carburant.",
    "he": "הזמן הטוב ביותר לעשות זאת הוא מיד לאחר שמילאת דלק."
  }
}

Built-In Language Analyzer Example

The following example index definition specifies an index on the subject.fr field using the french analyzer:

{
  "mappings": {
    "fields": {
      "subject": {
        "fields": {
          "fr": {
            "analyzer": "lucene.french",
            "type": "string"
          }
        },
        "type": "document"
      }
    }
  }
}

The following query searches for the string pour in the subject.fr field:

db.cars.aggregate([
  {
    $search: {
      "text": {
        "query": "pour",
        "path": "subject.fr"
      }
    }
  },
  {
    $project: {
      "_id": 0,
      "subject.fr": 1
    }
  }
])

The previous query returns no results when using the french analyzer, because pour is a built-in stop word. Using the standard analyzer, the same query would return both documents.

The following query searches for the string carburant in the subject.fr field:

db.cars.aggregate([
  {
    $search: {
      "text": {
        "query": "carburant",
        "path": "subject.fr"
      }
    }
  },
  {
    $project: {
      "_id": 0,
      "subject.fr": 1
    }
  }
])

{ subject: { fr: "Le meilleur moment pour le faire c'est immédiatement après que vous aurez fait le plein de carburant." } }

Atlas Search returns a document with _id: 1 in the results because the query matched a token that the lucene.french analyzer created for the document. The lucene.french analyzer creates the following tokens for the subject.fr field in document with _id: 1:

`meileu`	`moment`	`fair`
`est`	`imediat`	`aprè`
`fait`	`plein`	`carburant`

Custom Language Analyzer Example

You can also create indexes for unsupported languages by creating a custom analyzer with the icuFolding and stopword token filters.

The following example index definition specifies an index on the subject.he field using a custom analyzer called myHebrewAnalyzer to analyze and create tokens for Hebrew text:

{
  "analyzer": "lucene.standard",
  "mappings": {
    "dynamic": false,
    "fields": {
      "subject": {
        "fields": {
          "he": {
            "analyzer": "myHebrewAnalyzer",
            "type": "string"
          }
        },
        "type": "document"
      }
    }
  },
  "analyzers": [
    {
      "charFilters": [],
      "name": "myHebrewAnalyzer",
      "tokenFilters": [
        {
          "type": "icuFolding"
        },
        {
          "tokens": [
            "אן",
            "שלנו",
            "זה",
            "אל"
          ],
          "type": "stopword"
        }
      ],
      "tokenizer": {
        "type": "standard"
      }
    }
  ]
}

The following query searches for the string המכוניות in the subject.he field:

db.cars.aggregate([
  {
    $search: {
      "text": {
        "query": "המכוניות",
        "path": "subject.he"
      }
    }
  },
  {
    $project: {
      "_id": 0,
      "subject.he": 1
    }
  }
])

{ subject: { he: 'עדיף לצייד את המכוניות שלנו כדי להבין את הגורמים לתאונה.' } }

Atlas Search returns a document with _id: 1 in the results because the query matched a token that the myHebrewAnalyzer analyzer created for document. The myHebrewAnalyzer analyzer creates the following tokens for the subject.he field in document with _id: 1:

`עדיף`	`לצייד`	`את`
`המכוניות`	`כדי`	`להבין`
`את`	`הגורמים`	`לתאונה`

Back

Keyword

Multi