For AI agents: this page is also available as Markdown at /docs/arabic.md. Every docs page is: add .md to its URL. llms.txt lists them all.
Arabic
The arabic analyzer folds the spelling variants Arabic text mixes freely, removes common stop words and strips prefixes and suffixes (Lucene's light10 stemmer), so the definite article and plural endings do not stop a match: a search for كتاب finds الكتاب and كتابات.
{"fields": {"title_ar": {"type": "text", "analyzer": "arabic"}}}
The chain is unicode_word, nfkc_cf, arabic_normalize, the Lucene Arabic stop words and arabic_stem, the same as Elasticsearch's arabic analyzer, so an index migrated from Elasticsearch matches the same words. The stop words are normalized the same way as the text, so a stop word written with any alef variant is removed.
#Filters
The analyzer's two filters can be used in your own chain:
| Filter | Does |
|---|---|
arabic_normalize | Alef variants to bare alef, alef maksura to yeh, teh marbuta to heh; removes tatweel and harakat (vowel marks). |
arabic_stem | Removes common prefixes (ال, وال, بال, ...) and suffixes, keeping a minimum stem length. |
For example, to keep stop words (so phrases such as titles match in full) but still normalize and stem:
{
"type": "text",
"analyzer": {"tokenizer": "unicode_word", "filters": ["nfkc_cf", "arabic_normalize", "arabic_stem"]}
}
Prefix, wildcard and fuzzy terms are normalized with arabic_normalize but not stemmed.
#Through the compatible APIs
The Elasticsearch and OpenSearch analyzer arabic, the filters arabic_normalization and stemmer with language: arabic, and Solr's text_ar type and Arabic normalization and stem factories all map to these components. See Through the compatible APIs.