Elasticsearch plugin providing support for Serbian language.
Version of the plugin will follow version of the Elasticsearch it can be used with.
This plugin is used for preparing and analyzing text for indexing. It origins from this repo which was written based on following dependencies:
- Apache Lucene v4.9
- Elasticsearch v1.3.2
Since these versions are outdated, plugin codebase was rewritten and adapted to use these dependencies:
- Apache Lucene v7.2.1
- Elasticsearch v6.2.0
Unused, deprecated and redundant code is removed. Several parts are optimized, mainly with the use of Apache Commons Lang library.
SerbianAnalyzer uses following components during the analysis:
StandardTokenizerLowerCaseFilterLatCyrFilter(converts from cyrillic to latin letters)StopFilterSnowballFilterwith SerbianStemmer (stems tokens, taking only root of the word)RemoveAccentsFilter(remove accents from the letters)
To build plugin run the following command:
mvn clean install
To install build plugin to Elasticsearch run the following command:
elasticsearch-plugin install file:///path/to/plugin.zip
Analyzer provided by this plugin can afterwards be used in Elasticsearch environment by invocation via serbian-analyzer name.
Original stemmer written in Snowball string processing language is provided in the
resources/serbian-stemmber.sbl. However, with the latest Snowball compiler it can't be compiled without errors.
Because of that, already generated SerbianStemmer was just slightly adapted so it can be used with newer version of Lucene dependencies.