Hunspell token filter

edit

Provides dictionary stemming based on a provided Hunspell dictionary. The hunspell filter requires configuration of one or more language-specific Hunspell dictionaries.

This filter uses Lucene’s HunspellStemFilter.

If available, we recommend trying an algorithmic stemmer for your language before using the hunspell token filter. In practice, algorithmic stemmers typically outperform dictionary stemmers. See Dictionary stemmers.

Configure Hunspell dictionaries

edit

By default, Hunspell dictionaries are stored and detected on a dedicated hunspell directory on the filesystem: <path.config>/hunspell. Each dictionary is expected to have its own directory, named after its associated language and locale (e.g., pt_BR, en_GB). This dictionary directory is expected to hold a single .aff and one or more .dic files, all of which will automatically be picked up. For example, assuming the default <path.config>/hunspell path is used, the following directory layout will define the en_US dictionary:

- config
    |-- hunspell
    |    |-- en_US
    |    |    |-- en_US.dic
    |    |    |-- en_US.aff

Each dictionary can be configured with one setting:

ignore_case
(Static, boolean) If true, dictionary matching will be case insensitive. Defaults to false.

This setting can be configured globally in elasticsearch.yml using indices.analysis.hunspell.dictionary.ignore_case.

To configure the setting for a specific locale, use the indices.analysis.hunspell.dictionary.<locale>.ignore_case setting (e.g., for the en_US (American English) locale, the setting is indices.analysis.hunspell.dictionary.en_US.ignore_case).

It is also possible to add settings.yml file under the dictionary directory which holds these settings. This overrides any other ignore_case settings defined in elasticsearch.yml.

Example

edit

The following analyze API request uses the hunspell filter to stem the foxes jumping quickly to the fox jump quick.

The request specifies the en_US locale, meaning that the .aff and .dic files in the <path.config>/hunspell/en_US directory are used for the Hunspell dictionary.

GET /_analyze
{
  "tokenizer": "standard",
  "filter": [
    {
      "type": "hunspell",
      "locale": "en_US"
    }
  ],
  "text": "the foxes jumping quickly"
}

The filter produces the following tokens:

[ the, fox, jump, quick ]

Configurable parameters

edit
dictionary

(Optional, string or array of strings) One or more .dic files (e.g, en_US.dic, my_custom.dic) to use for the Hunspell dictionary.

By default, the hunspell filter uses all .dic files in the <path.config>/hunspell/<locale> directory specified specified using the lang, language, or locale parameter. To use another directory, the directory’s path must be registered using the indices.analysis.hunspell.dictionary.location setting.

dedup
(Optional, boolean) If true, duplicate tokens are removed from the filter’s output. Defaults to true.
lang

(Required*, string) An alias for the locale parameter.

If this parameter is not specified, the language or locale parameter is required.

language

(Required*, string) An alias for the locale parameter.

If this parameter is not specified, the lang or locale parameter is required.

locale

(Required*, string) Locale directory used to specify the .aff and .dic files for a Hunspell dictionary. See Configure Hunspell dictionaries.

If this parameter is not specified, the lang or language parameter is required.

longest_only
(Optional, boolean) If true, only the longest stemmed version of each token is included in the output. If false, all stemmed versions of the token are included. Defaults to false.

Customize and add to an analyzer

edit

To customize the hunspell filter, duplicate it to create the basis for a new custom token filter. You can modify the filter using its configurable parameters.

For example, the following create index API request uses a custom hunspell filter, my_en_US_dict_stemmer, to configure a new custom analyzer.

The my_en_US_dict_stemmer filter uses a locale of en_US, meaning that the .aff and .dic files in the <path.config>/hunspell/en_US directory are used. The filter also includes a dedup argument of false, meaning that duplicate tokens added from the dictionary are not removed from the filter’s output.

PUT /my-index-000001
{
  "settings": {
    "analysis": {
      "analyzer": {
        "en": {
          "tokenizer": "standard",
          "filter": [ "my_en_US_dict_stemmer" ]
        }
      },
      "filter": {
        "my_en_US_dict_stemmer": {
          "type": "hunspell",
          "locale": "en_US",
          "dedup": false
        }
      }
    }
  }
}

Settings

edit

In addition to the ignore_case settings, you can configure the following global settings for the hunspell filter using elasticsearch.yml:

indices.analysis.hunspell.dictionary.lazy
(Static, boolean) If true, the loading of Hunspell dictionaries is deferred until a dictionary is used. If false, the dictionary directory is checked for dictionaries when the node starts, and any dictionaries are automatically loaded. Defaults to false.
indices.analysis.hunspell.dictionary.location

(Static, string) Path to a Hunspell dictionary directory. This path must be absolute or relative to the config location.

By default, the <path.config>/hunspell directory is used, as described in Configure Hunspell dictionaries.