Skip to content
develop
Switch branches/tags
Code

Latest commit

 

Git stats

Files

Permalink
Failed to load latest commit information.
Type
Name
Latest commit message
Commit time
 
 
src
 
 
 
 
 
 
 
 
 
 
 
 
 
 

analysis-sudachi

analysis-sudachi is an Elasticsearch plugin for tokenization of Japanese text using Sudachi the Japanese morphological analyzer.

build Quality Gate Status

What's new?

  • version 2.1.0

    • Added a new property additional_settings to write Sudachi settings directly in config
    • Added support for specifying Elasticsearch version at build time
  • version 2.0.3

    • Fix duplicated tokens for OOVs with sudachi_split filter's extended mode
  • version 2.0.2

    • Upgrade Sudachi to 0.4.3
      • Fix overrun with surrogate pairs
  • version 2.0.1

    • Upgrade Sudachi to 0.4.2
      • Fix buffer overrun with character normalization
  • version 2.0.0

    • New mode split_mode was added
    • New filter sudachi_split was added instead of mode
    • mode was deperecated
    • Upgrade Sudachi morphological analyzer to 0.4.1
    • Words containing periods are no longer split
    • Fix a bug causing wrong offsets with icu_normalizer
  • version 1.3.2

    • Upgrade Sudachi morphological analyzer to 0.3.1
  • version 1.3.1

    • Upgrade Sudachi morphological analyzer to 0.3.0
    • Minor bug fix
  • version 1.3.0

    • Upgrade Sudachi morphological analyzer to 0.2.0
    • Import Sudachi from maven central repository
    • Minor bug fix
  • version 1.2.0

    • Upgrading Sudachi morphological analyzer to 0.2.0-SNAPSHOT
    • New filter sudachi_normalizedform was added; see sudachi_normalizedform
    • Default normalization behavior was changed; neather baseform filter and normalziedform filter not applied
    • sudachi_readingform filter was changed with new romaji mappings based on MS-IME
  • version 1.1.0

  • version 1.0.0

    • first release

Build

  1. Build analysis-sudachi.
   $ ./gradlew -PelasticsearchVersion=7.10.1 build
  • develop branch: for Elasticsearch 7.8 or later
  • es7.4-7.7 branch: for Elasticsearch 7.4, 7.5, 7.6, 7.7
  • es7.0-7.3 branch: for Elasticsearch 7.0, 7.1, 7.2, 7.3
  • es6.8 branch: for Elasticsearch 5.8
  • es5.6 branch: for Elasticsearch 5.6

Installation

  1. Download analysis-sudachi-elasticsearch zip archive file
  2. Move current dir to $ES_HOME
  3. Execute "bin/elasticsearch-plugin install file:///plugin-zip-path"
  4. Download sudachi dictionary archive from https://github.com/WorksApplications/SudachiDict
  5. Extract dic file and place it to config/sudachi/system_core.dic (You must install system_core.dic in this place if you use Elasticsearch 7.6 or later)
  6. Execute "bin/elasticsearch"

Configuration

  • split_mode: Select splitting mode of Sudachi. (A, B, C) (string, default: C)
    • C: Extracts named entities
      • Ex) 選挙管理委員会
    • B: Into the middle units
      • Ex) 選挙,管理,委員会
    • A: The shortest units equivalent to the UniDic short unit
      • Ex) 選挙,管理,委員,会
  • discard_punctuation: Select to discard punctuation or not. (bool, default: true)
  • settings_path: Sudachi setting file path. The path may be absolute or relative; relative paths are resolved with respect to es_config. (string, default: null)
  • resources_path: Sudachi dictionary path. The path may be absolute or relative; relative paths are resolved with respect to es_config. (string, default: null)
  • additional_settings: Describes a configuration JSON string for Sudachi. This JSON string will be merged into the default configuration. If this property is set, settings_path will be ignored.

Example

{
  "settings": {
    "index": {
      "analysis": {
        "tokenizer": {
          "sudachi_tokenizer": {
            "type": "sudachi_tokenizer",
            "split_mode": "C",
            "discard_punctuation": true,
            "resources_path": "/etc/elasticsearch/sudachi"
          }
        },
        "analyzer": {
          "sudachi_analyzer": {
            "filter": [],
            "tokenizer": "sudachi_tokenizer",
            "type": "custom"
          }
        }
      }
    }
  }
}

Dictionary

You can specify the dictionary either in the file specified by settings_path or by additional_settings.

Example

{
  "settings": {
    "index": {
      "analysis": {
        "tokenizer": {
          "sudachi_tokenizer": {
            "type": "sudachi_tokenizer",
            "additional_settings": "{\"systemDict\":\"system_full.dic\",\"userDict\":[\"user.dic\"]}"
          }
        },
        "analyzer": {
          "sudachi_analyzer": {
            "filter": [],
            "tokenizer": "sudachi_tokenizer",
            "type": "custom"
          }
        }
      }
    }
  }
}

Filters

sudachi_split

This filter works like mode of kuromoji.

  • search: Additional segmentation useful for search. (Use C and A mode)
    • Ex)関西国際空港, 関西, 国際, 空港 / アバラカダブラ
  • extended: Similar to search mode, but also unigram unknown words.
    • Ex)関西国際空港, 関西, 国際, 空港 / アバラカダブラ, ア, バ, ラ, カ, ダ, ブ, ラ

PUT sudachi_sample

{
  "settings": {
    "index": {
      "analysis": {
        "tokenizer": {
          "sudachi_tokenizer": {
            "type": "sudachi_tokenizer"
          }
        },
        "analyzer": {
          "sudachi_analyzer": {
            "filter": ["my_searchfilter" ],
            "tokenizer": "sudachi_tokenizer",
            "type": "custom"
          }
        },
        "filter":{
          "my_searchfilter": {
            "type": "sudachi_split",
            "mode": "search"
          }
        }
      }
    }
  }
}

POST sudachi_sample

{
    "analyzer": "sudachi_analyzer",
    "text": "関西国際空港"
}

Which responds with:

{
  "tokens" : [
    {
      "token" : "関西国際空港",
      "start_offset" : 0,
      "end_offset" : 6,
      "type" : "word",
      "position" : 0,
      "positionLength" : 3
    },
    {
      "token" : "関西",
      "start_offset" : 0,
      "end_offset" : 2,
      "type" : "word",
      "position" : 0
    },
    {
      "token" : "国際",
      "start_offset" : 2,
      "end_offset" : 4,
      "type" : "word",
      "position" : 1
    },
    {
      "token" : "空港",
      "start_offset" : 4,
      "end_offset" : 6,
      "type" : "word",
      "position" : 2
    }
  ]
}

sudachi_part_of_speech

The sudachi_part_of_speech token filter removes tokens that match a set of part-of-speech tags. It accepts the following setting:

The stopatgs is an array of part-of-speech and/or inflection tags that should be removed. It defaults to the stoptags.txt file embedded in the lucene-analysis-sudachi.jar.

Sudachi POS information is a csv list, consisting 6 items;

  • 1-4 part-of-speech hierarchy (品詞階層)
  • 5 inflectional type (活用型)
  • 6 inflectional form (活用形)

With the stoptags, you can filter out the result in any of these forward matching forms;

  • 1 - e.g., 名詞
  • 1,2 - e.g., 名詞,固有名詞
  • 1,2,3 - e.g., 名詞,固有名詞,地名
  • 1,2,3,4 - e.g., 名詞,固有名詞,地名,一般
  • 5 - e.g., 五段-カ行
  • 6 - e.g., 終止形-一般
  • 5,6 - e.g., 五段-カ行,終止形-一般

PUT sudachi_sample

{
  "settings": {
    "index": {
      "analysis": {
        "tokenizer": {
          "sudachi_tokenizer": {
            "type": "sudachi_tokenizer"
          }
        },
        "analyzer": {
          "sudachi_analyzer": {
            "filter": [ "my_posfilter" ],
            "tokenizer": "sudachi_tokenizer",
            "type": "custom"
          }
        },
        "filter":{
          "my_posfilter":{
            "type":"sudachi_part_of_speech",
            "stoptags":[
              "助詞",
              "助動詞",
              "補助記号,句点",
              "補助記号,読点"
            ]
          }
        }
      }
    }
  }
}

POST sudachi_sample

{
  "analyzer": "sudachi_analyzer",
  "text": "寿司がおいしいね"
}

Which responds with:

{
  "tokens": [
    {
      "token": "寿司",
      "start_offset": 0,
      "end_offset": 2,
      "type": "word",
      "position": 0
    },
    {
      "token": "美味しい",
      "start_offset": 3,
      "end_offset": 7,
      "type": "word",
      "position": 2
    }
  ]
}

sudachi_ja_stop

The sudachi_ja_stop token filter filters out Japanese stopwords (japanese), and any other custom stopwords specified by the user. This filter only supports the predefined japanese stopwords list. If you want to use a different predefined list, then use the stop token filter instead.

PUT sudachi_sample

{
  "settings": {
    "index": {
      "analysis": {
        "tokenizer": {
          "sudachi_tokenizer": {
            "type": "sudachi_tokenizer"
          }
        },
        "analyzer": {
          "sudachi_analyzer": {
            "filter": [ "my_stopfilter" ],
            "tokenizer": "sudachi_tokenizer",
            "type": "custom"
          }
        },
        "filter":{
          "my_stopfilter":{
            "type":"sudachi_ja_stop",
            "stopwords":[
              "_japanese_",
              "",
              "です"
            ]
          }
        }
      }
    }
  }
}

POST sudachi_sample

{
  "analyzer": "sudachi_analyzer",
  "text": "私は宇宙人です。"
}

Which responds with:

{
  "tokens": [
    {
      "token": "",
      "start_offset": 0,
      "end_offset": 1,
      "type": "word",
      "position": 0
    },
    {
      "token": "宇宙",
      "start_offset": 2,
      "end_offset": 4,
      "type": "word",
      "position": 2
    },
    {
      "token": "",
      "start_offset": 4,
      "end_offset": 5,
      "type": "word",
      "position": 3
    }
  ]
}

sudachi_baseform

The sudachi_baseform token filter replaces terms with their SudachiBaseFormAttribute. This acts as a lemmatizer for verbs and adjectives.

PUT sudachi_sample

{
  "settings": {
    "index": {
      "analysis": {
        "tokenizer": {
          "sudachi_tokenizer": {
            "type": "sudachi_tokenizer"
          }
        },
        "analyzer": {
          "sudachi_analyzer": {
            "filter": [ "sudachi_baseform" ],
            "tokenizer": "sudachi_tokenizer",
            "type": "custom"
          }
        }
      }
    }
  }
}

POST sudachi_sample

{
  "analyzer": "sudachi_analyzer",
  "text": "飲み"
}

Which responds with:

{
  "tokens": [
    {
      "token": "飲む",
      "start_offset": 0,
      "end_offset": 2,
      "type": "word",
      "position": 0
    }
  ]
}

sudachi_normalizedform

The sudachi_normalizedform token filter replaces terms with their SudachiNormalizedFormAttribute. This acts as a normalizer for spelling variants.

This filter lemmatizes verbs and adjectives too. You don't need to use sudachi_baseform filter with this filter.

PUT sudachi_sample

{
  "settings": {
    "index": {
      "analysis": {
        "tokenizer": {
          "sudachi_tokenizer": {
            "type": "sudachi_tokenizer"
          }
        },
        "analyzer": {
          "sudachi_analyzer": {
            "filter": [ "sudachi_normalizedform" ],
            "tokenizer": "sudachi_tokenizer",
            "type": "custom"
          }
        }
      }
    }
  }
}

POST sudachi_sample

{
  "analyzer": "sudachi_analyzer",
  "text": "呑み"
}

Which responds with:

{
  "tokens": [
    {
      "token": "飲む",
      "start_offset": 0,
      "end_offset": 2,
      "type": "word",
      "position": 0
    }
  ]
}

sudachi_readingform

Convert to katakana or romaji reading. The sudachi_readingform token filter replaces the token with its reading form in either katakana or romaji. It accepts the following setting:

use_romaji

Whether romaji reading form should be output instead of katakana. Defaults to false.

When using the pre-defined sudachi_readingform filter, use_romaji is set to true. The default when defining a custom sudachi_readingform, however, is false. The only reason to use the custom form is if you need the katakana reading form:

PUT sudachi_sample

{
  "settings": {
    "index": {
      "analysis": {
        "filter": {
          "romaji_readingform": {
            "type": "sudachi_readingform",
            "use_romaji": true
          },
          "katakana_readingform": {
            "type": "sudachi_readingform",
            "use_romaji": false
          }
        },
        "tokenizer": {
          "sudachi_tokenizer": {
            "type": "sudachi_tokenizer"
          }
        },
        "analyzer": {
          "romaji_analyzer": {
            "tokenizer": "sudachi_tokenizer",
            "filter": [ "romaji_readingform" ]
          },
          "katakana_analyzer": {
            "tokenizer": "sudachi_tokenizer",
            "filter": [ "katakana_readingform" ]
          }
        }
      }
    }
  }
}

POST sudachi_sample

{
  "analyzer": "katakana_analyzer",
  "text": "寿司"
}

Returns スシ.

{
  "analyzer": "romaji_analyzer",
  "text": "寿司"
}

Returns susi.

License

Copyright (c) 2017-2020 Works Applications Co., Ltd. Originally under elasticsearch, https://www.elastic.co/jp/products/elasticsearch Originally under lucene, https://lucene.apache.org/