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.
MongoDB Search offers the following language analyzers:
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
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 MongoDB Search query searches for the string pour in the subject.fr field. To run this query, connect to your cluster using mongosh and switch to the database that contains the cars collection.
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 MongoDB Search query searches for the string carburant in the subject.fr field. To run this query, connect to your cluster using mongosh and switch to the database that contains the cars collection.
db.cars.aggregate([ { $search: { "text": { "query": "carburant", "path": "subject.fr" } } }, { $project: { "_id": 0, "subject.fr": 1 } } ])
MongoDB 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:
|
|
|
|
|
|
|
|
|
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 MongoDB Search query searches for the string המכוניות in the subject.he field. To run this query, connect to your cluster using mongosh and switch to the database that contains the cars collection.
db.cars.aggregate([ { $search: { "text": { "query": "המכוניות", "path": "subject.he" } } }, { $project: { "_id": 0, "subject.he": 1 } } ])
MongoDB 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:
|
|
|
|
|
|
|
|
|
Multilingual Search Example
You can also create an index that uses multiple language analyzers to perform a multilingual search.
The following example index definition specifies an index with dynamic mapping on the sample_mflix.movies collection. The definition applies the lucene.italian language analyzer to index the fullplot field, and uses the multi option to specify lucene.english as an alternate language analyzer. MongoDB Search uses the default lucene.english language analyzer for all other fields that it dynamically indexes in the movies collection.
{ "analyzer": "lucene.standard", "mappings": { "dynamic": true, "fields": { "fullplot": { "type": "string", "analyzer": "lucene.italian", "multi": { "fullplot_english": { "type": "string", "analyzer": "lucene.english", } } } } } }
The following MongoDB Search query uses the compound operator to query the collection in multiple languages. To run this query, connect to your cluster using mongosh and switch to the sample_mflix database.
The compound operator contains the following clauses:
mustclause searches for movie plots in English and Italian that contain the termBellausing the text operatormustNotclause excludes movies released between the years 1984 to 2016 using the range operatorshouldclause specifies preference for theComedygenre using the text operator
db.movies.aggregate([ { $search: { "index": "multilingual-tutorial", "compound": { "must": [{ "text": { "query": "Bella", "path": { "value": "fullplot", "multi": "fullplot_english" } } }], "mustNot": [{ "range": { "path": "released", "gt": ISODate("1984-01-01T00:00:00.000Z"), "lt": ISODate("2016-01-01T00:00:00.000Z") } }], "should": [{ "text": { "query": "Comedy", "path": "genres" } }] } } }, { $project: { "_id": 0, "title": 1, "plot": 1, "genres": 1, "runtime": 1, "fullplot": 1, "released": 1, "score": { "$meta": "searchScore" } } } ])