Taming Text


Just beginning ‘Taming Text – How to find, organize and manipulate it’

  1. Getting started and taming text
  2. Foundations of taming text
  3. Searching
  4. Fuzzy string matching
  5. Identifying people, places and things
  6. Clustering text
  7. Classification, categorization and tagging
  8. Building an example question answering system
  9. Untamed text: exploring the next frontier


Model Data to Support Keyword Search

  • Data Models > 
  • Model Data to Support Keyword Search


MongoDB 2.4 text searc




New in version 2.4.

Searches text content stored in the text index. The text command is case-insensitive.

The text command returns all documents that contain any of the terms; i.e. it performs a logical OR search. By default, the command limits the matches to the top 100 scoring documents, in descending score order, but you can specify a different limit.

The text command has the following syntax:

db.collection.runCommand( "text", { search: <string>,
                                    filter: <document>,
                                    project: <document>,
                                    limit: <number>,
                                    language: <string> } )

The text command has the following parameters:

Field Type Description
search string A string of terms that MongoDB parses and uses to query the text index. Enclose the string of terms in escaped double quotes to match on the phrase. For further information on the search field syntax, see The search Field.
filter document Optional. A query document to further limit the results of the query using another database field. Use any valid MongoDB query in the filter document, except if the index includes an ascending or descending index field as a prefix. If the index includes an ascending or descending index field as a prefix, the filter is required and the filter query must be an equality match.
project document Optional. Limits the fields returned by the query to only those specified. By default, the _id field returns as part of the result set, unless you explicitly exclude the field in the project document.
limit number Optional. The maximum number of documents to include in the response. The textcommand sorts the results before applying the limit. The default limit is 100.
language string Optional. The language that determines the list of stop words for the search and the rules for the stemmer and tokenizer. If not specified, the search uses the default language of the index. For supported languages, see Text Search Languages. Specify the language in lowercase.
Returns: The text command returns a document that contains a field results that contains an array of the highest scoring documents, in descending order by score. See Output for details.



The complete results of the text command must fit within the BSON Document Size. Otherwise, the command will limit the results to fit within the BSON Document Size. Use the limit and the project parameters with the text command to limit the size of the result set.


  • If the search string includes phrases, the search performs an AND with any other terms in the search string; e.g. search for"\"twinkle twinkle\" little star" searches for "twinkle twinkle" and ("little" or "star").
  • text adds all negations to the query with the logical AND operator.
  • The text command ignores stop words for the search language, such as the and and in English.
  • The text command matches on the complete stemmed word. So if a document field contains the word blueberry, a search on the term blue will not match. However, blueberry or blueberries will match.



You cannot combine the text command, which requires a special text index, with a query operator that requires a different type of special index. For example you cannot combine text with the $near operator.

The search Field

The search field takes a string of terms that MongoDB parses and uses to query the text index. Enclose the string of terms in escaped double quotes to match on the phrase. Additionally, the text command treats most punctuation as delimiters, except when a hyphen - negates terms.

Prefixing a word with a hyphen sign (-) negates a word:

  • The negated word excludes documents that contain the negated word from the result set.
  • A search string that only contains negated words returns no match.
  • A hyphenated word, such as pre-market, is not a negation. The text command treats the hyphen as a delimiter.


The following examples assume a collection articles that has a text index on the field subject:

db.articles.ensureIndex( { subject: "text" } )

Search for a Single Word

db.articles.runCommand( "text", { search: "coffee" } )

This query returns documents that contain the word coffee, case-insensitive, in the indexed subject field.

Search for Multiple Words

The following command searches for bake or coffee or cake:

db.articles.runCommand( "text", { search: "bake coffee cake" } )

This query returns documents that contain either bake or coffee or cake in the indexed subject field.

Search for a Phrase

db.articles.runCommand( "text", { search: "\"bake coffee cake\"" } )

This query returns documents that contain the phrase bake coffee cake.

Exclude a Term from the Result Set

Use the hyphen (-) as a prefix to exclude documents that contain a term. Search for documents that contain the words bake or coffee but do not contain cake:

db.articles.runCommand( "text", { search: "bake coffee -cake" } )

Search with Additional Query Conditions

Use the filter option to include additional query conditions.

Search for a single word coffee with an additional filter on the about field, but limit the results to 2 documents with the highest score and return only the subject field in the matching documents:

db.articles.runCommand( "text", {
                                  search: "coffee",
                                  filter: { about: /desserts/ },
                                  limit: 2,
                                  project: { subject: 1, _id: 0 }
  • The filter query document may use any of the available query operators.
  • Because the _id field is implicitly included, in order to return only the subject field, you must explicitly exclude (0) the _id field. Within the project document, you cannot mix inclusions (i.e. <fieldA>: 1) and exclusions (i.e. <fieldB>: 0), except for the_id field.

Search a Different Language

Use the language option to specify Spanish as the language that determines the list of stop words and the rules for the stemmer and tokenizer:

db.articles.runCommand( "text", {
                                    search: "leche",
                                    language: "spanish"

See Text Search Languages for the supported languages.



Specify the language in lowercase.


The following is an example document returned by the text command:

   "queryDebugString" : "tomorrow||||||",
   "language" : "english",
   "results" : [
         "score" : 1.3125,
         "obj": {
                  "_id" : ObjectId("50ecef5f8abea0fda30ceab3"),
                  "quote" : "tomorrow, and tomorrow, and tomorrow, creeps in this petty pace",
                  "related_quotes" : [
                                       "is this a dagger which I see before me",
                                       "the handle toward my hand?"
                  "src" : {
                             "title" : "Macbeth",
                             "from" : "Act V, Scene V"
                  "speaker" : "macbeth"
   "stats" : {
               "nscanned" : 1,
               "nscannedObjects" : 0,
               "n" : 1,
               "nfound" : 1,
               "timeMicros" : 163
   "ok" : 1

The text command returns the following data:

For internal use only.

The language field returns the language used for the text search. This language determines the list of stop words and the rules for the stemmer and tokenizer.

The results field returns an array of result documents that contain the information on the matching documents. The result documents are ordered by the score. Each result document contains:

The obj field returns the actual document from the collection that contained the stemmed term or terms.

The score field for the document that contained the stemmed term or terms. The score field signifies how well the document matched the stemmed term or terms. See Control Results of Text Search with Weights for how you can adjust the scores for the matching words.

The stats field returns a document that contains the query execution statistics. The stats field contains:

The nscanned field returns the total number of index entries scanned.

The nscannedObjects field returns the total number of documents scanned.

The n field returns the number of elements in the results array. This number may be less than the total number of matching documents, i.e. nfound, if the full result exceeds the BSON Document Size.

The nfound field returns the total number of documents that match. This number may be greater than the size of the results array, i.e. n, if the result set exceeds the BSON Document Size.

The timeMicros field returns the time in microseconds for the search.

The ok returns the status of the text command.

Text Search Languages

The text index and the text command support the following languages:

  • danish
  • dutch
  • english
  • finnish
  • french
  • german
  • hungarian
  • italian
  • norwegian
  • portuguese
  • romanian
  • russian
  • spanish
  • swedish
  • turkish



If you specify a language value of "none", then the text search has no list of stop words, and the text search does not stem or tokenize the search terms.

MongoDB text analytics

Version 2.4 of MongoDB (seems to be getting richer all the time, if only they would cut a Raspberry Pi version!) will be released with full text search capabilities, which should mean text mining tweets etc as blogged here might become a whole lot easier natively, without needing to do stuff in R too? The functionality includes stop word identification and stemming.