This is a custom Phonetic Analysis plugin for brazilian portuguese using MTFN algorithm - an implementation of Metaphone for pt-BR (and mostly compatible with pt-PT).

• JDK 1.8
• Maven 3.x
• Elasticsearch 6.2.1

First of all, make sure that Maven can use MTFN as a dependency. At this time, it is not available in Maven Central, so you can add it to your private repository or install it in your local machine.

git clone https://github.com/ruliana/MTFN.git
cd MTFN
mvn clean install

Then building this plugin is easy as pie.

mvn clean package

You will find the installation package at ./target/releases/.

This plugin can be installed using the plugin manager, in offline installation mode. To install analysis-mtfn from your local file system at /path/to/analysis-mtfn.zip:

sudo bin/elasticsearch-plugin install file:https:///path/to/analysis-mtfn.zip

Find more information about Elasticsearch plugin offline installation at Plugin Management » Custom URL or file system.

The plugin can be removed with the following command:

sudo bin/elasticsearch-plugin remove analysis-mtfn

You can configure your index to support phonetic searches in brazilian portuguese by adding this plugin as a filter with type phonetic_br to your index settings, defining a new analyzer and binding it to a field mapping.

{
  "settings": {
    "analysis": {
      "filter": {
        "brazilian_phonetic": {
          "type": "phonetic_br"
        }
      },
      "analyzer": {
        "brazilian_phonetic": {
          "type": "custom",
          "tokenizer": "standard",
          "filter": [
            "brazilian_phonetic"
          ]
        }
      }
    }
  },
  "mappings": {
    "properties": {
      "nome": {
        "type": "text",
        "fields": {
          "phonetic": {
            "type": "text",
            "index": true,
            "analyzer": "brazilian_phonetic"
          }
        }
      }
    }
  }
}

To be defined