Khmerlang Analysis Plugin for Elasticsearch

Khmerlang Analysis plugin integrates Khmer language analysis into Elasticsearch.

The plugin provides kh_analyzer analyzer, kh_tokenizer tokenizer, kh_synonym khmer synonym, kh_number khmer number filter and kh_stop stop filter. The kh_analyzer is composed of the kh_tokenizer tokenizer, correct_character, kh_number, stop and lowercase filter.

• Re-order character: sometime words look correct however the position of characters is not. Ex “ស្រី្ត” can written as "ស + ្ + រ + ្ + ត + ី" or "ស + ្ + ត + ្ + រ + ី".
• Correct character: sometime words look correct however the character using is not correct. Ex: "ប្ដី", correct: "ប + ្ + ដ + ី", incorrect: "ប + ្ + ត + ី".
• Word Segmentation: build token from to segment words of input text.
• Synonyms: add synonym for some token.

Example output

GET _analyze
{
  "analyzer": "kh_analyzer",
  "text": "ខ្ញុំស្រលាញ់កម្ពុជា។"
}

The above sentence would produce the following terms:

{
  "tokens" : [
    {
      "token" : "ស្រលាញ់",
      "start_offset" : 5,
      "end_offset" : 11,
      "type" : "<KH>",
      "position" : 1
    },
    {
      "token" : "កម្ពុជា",
      "start_offset" : 12,
      "end_offset" : 18,
      "type" : "<KH>",
      "position" : 2
    }
  ]
}

Configuration

The kh_analyzer analyzer accepts the following parameters:

• correct_character Correct character order. Defaults to true.
• khmer_number convert khmer number to arabic. Defaults to false.
• lowercase Convert character to lowercase. Defaults to false.
• dict_path The path to tokenizer dictionary on system(TODO: not test yet).
• keep_punctuation Keep punctuation marks as tokens. Defaults to false.
• enable_stopwords Enable/disable stop words filter. Defaults to false.
• stopwords A pre-defined stop words list. Defaults to stopwords.txt file.
• stopwords_path The path to a file containing stop words.

Example configuration

In this example, we configure the kh_analyzer analyzer to keep punctuation marks and to use the custom list of stop words:

PUT my-kh-index-00001
{
  "settings": {
    "analysis": {
      "analyzer": {
        "my_kh_analyzer": {
          "type": "kh_analyzer",
          "keep_punctuation": false,
          "correct_character": true,
          "lowercase": true,
          "khmer_number": true
        }
      }
    }
  }
}

GET my-kh-index-00001/_analyze
{
  "analyzer": "my_kh_analyzer",
  "text": "១២៣៤៥.៦៧អ្នកចេះ​និយាយភាសាខ្មែរទេ? 12345.67"
}

The above example produces the following terms:

{
  "tokens" : [
    {
      "token" : "12345.67",
      "start_offset" : 0,
      "end_offset" : 7,
      "type" : "<NUMBER>",
      "position" : 0
    },
    {
      "token" : "អ្នកចេះ",
      "start_offset" : 8,
      "end_offset" : 14,
      "type" : "<KH_WORD>",
      "position" : 1
    },
    {
      "token" : "និយាយ",
      "start_offset" : 15,
      "end_offset" : 19,
      "type" : "<KH_WORD>",
      "position" : 2
    },
    {
      "token" : "ភាសាខ្មែរ",
      "start_offset" : 20,
      "end_offset" : 28,
      "type" : "<KH_WORD>",
      "position" : 3
    },
    {
      "token" : "12345.67",
      "start_offset" : 34,
      "end_offset" : 41,
      "type" : "<NUMBER>",
      "position" : 5
    }
  ]
}

We can also create a custom analyzer with the kh_tokenizer. In following example, we create my_kh_analyzer to produce both diacritic and no diacritic tokens in lowercase:

PUT my-kh-index-00002
{
  "settings": {
    "analysis": {
      "analyzer": {
        "my_kh_analyzer": {
          "tokenizer": "kh_tokenizer",
          "filter": [
            "kh_synonym",
            "kh_number",
            "kh_stop",
            "lowercase"
          ]
        }
      }
    }
  }
}

GET my-kh-index-00002/_analyze
{
  "analyzer": "my_kh_analyzer",
  "text": "១២៣៤៥.៦៧អ្នកចេះ​និយាយភាសាខ្មែរទេ? 1234"
}

The above example produces the following terms:

{
  "tokens" : [
    {
      "token" : "12345.67",
      "start_offset" : 0,
      "end_offset" : 7,
      "type" : "<NUMBER>",
      "position" : 0
    },
    {
      "token" : "អ្នកចេះ",
      "start_offset" : 8,
      "end_offset" : 14,
      "type" : "<KH_WORD>",
      "position" : 1
    },
    {
      "token" : "និយាយ",
      "start_offset" : 15,
      "end_offset" : 19,
      "type" : "<KH_WORD>",
      "position" : 2
    },
    {
      "token" : "ភាសាខ្មែរ",
      "start_offset" : 20,
      "end_offset" : 28,
      "type" : "<KH_WORD>",
      "position" : 3
    },
    {
      "token" : "1234",
      "start_offset" : 34,
      "end_offset" : 37,
      "type" : "<NUMBER>",
      "position" : 5
    }
  ]
}

TODO

• word segment using deeplearning
• build synonym word list, group by category

Use Docker

Make sure you have installed both Docker & docker-compose

Build the image with Docker Compose

# Copy, edit ES version and password for user elastic in file .env. Default password: changeme
cp .env.sample .env
docker compose build
docker compose up

Step 1: Build the plugin

Clone the plugin’s source code:

git clone git@github.com:khmerlang/elasticsearch-analysis-khmerlang.git

Optionally, edit the elasticsearch-analysis-khmerlang/pom.xml to change the version of Elasticsearch (same as plugin version) you want to build the plugin with:

...
<version>7.17.10</version>
...

Build the plugin:

cd elasticsearch-analysis-khmerlang
mvn package

Step 2: Installation the plugin on Elasticsearch

bin/elasticsearch-plugin install file://target/releases/elasticsearch-analysis-khmerlang-7.17.1.zip

Khmerlang Analysis Plugin	Elasticsearch
master	7.17.5-7.17.10
develop	7.17.5-7.17.10
7.17.5-7.17.10	7.17.5-7.17.10
7.16.1	7.16 ~ 7.17.1
...	...

Issues:

• Build fail due to Java version in correct. Change java version to 11: export JAVA_HOME=$(/usr/libexec/java_home -v 11)

Refs:

• https://www.elastic.co/blog/multitoken-synonyms-and-graph-queries-in-elasticsearch

• https://www.elastic.co/blog/boosting-the-power-of-elasticsearch-with-synonyms

• https://www.elastic.co/guide/en/elasticsearch/plugins/8.3/analysis-kuromoji.html

• https://medium.com/@purbon/handling-similar-words-in-elasticsearch-9c80aba88627

• https://github.com/agora-team/elasticsearch-synonyms

• https://github.com/bells/elasticsearch-analysis-dynamic-synonym

• https://lucidworks.com/post/search-automatic-synonym-detection/

• https://github.com/jianfengye/elasticsearch-synonym-remote

• https://github.com/duydo/elasticsearch-analysis-vietnamese

• https://github.com/apache/lucene

• https://nlp.techostartup.center/