Supported languages for indexing features

This is for:

In this article

Troubleshooting language indexing issues

Language support and relevance disclaimer

The Coveo Platform supports a wide range of languages across indexing, machine learning models, and user interfaces; however, the depth of linguistic processing varies by language. Support for a language doesn’t constitute a commitment to equivalent relevance, accuracy, or feature parity.

For example, queries and indexed content in languages such as English, French, German, and Spanish achieve higher and more consistent relevance due to more comprehensive language analysis and optimized platform components. Conversely, for other languages that are technically supported but for which the platform isn’t fully optimized, search relevance, generated answers, and overall accuracy can be reduced or be less consistent. Accordingly, language support doesn’t guarantee equal relevance quality, and observed outcomes can vary depending on the language.

The table below lists the 62 languages Coveo supports where encoding, excerpt, language detection, and thesaurus features are available for indexing. The index recognizes any ISO 639-1 language code. For the languages listed in the table below, the index applies dedicated language-specific processing, such as stemming and language detection. For any other ISO 639-1 language code, the index recognizes it but doesn’t apply dedicated language processing. Coveo can also index content in such languages as long as:

The language uses spaces to separate words.
The item encodes characters in Unicode.

For languages meeting these requirements, most language features are supported except language detection, decompounding, and stemming.

Note

Since result ranking is partly based on summarization and key-concept extraction techniques, English, French, German, and Spanish queries return the most relevant results consistently.

Language Locale Encoding, excerpt, and thesaurus Did you mean Language detection Stemming Decompounding

English^[1]

en

N/A

French^[1]

fr

N/A

German^[1]

de

Spanish^[1]

es

N/A

Danish

da

Dutch

nl

Finnish

fi

Hungarian

hu

N/A

Italian

it

N/A

Norwegian

no

Portuguese

pt

N/A

Swedish

sv

Turkish

tr

N/A

Catalan

ca

N/A

Romanian

ro

N/A

Valencian

ca

N/A

Armenian

hy

N/A

Russian^[5]

ru

N/A

Chinese (traditional and simplified)^[2],[3]

zh

N/A

Greek

el

N/A

Hindi^[5]

hi

N/A

Japanese^[2],[3]

ja

N/A

Korean^[2],[3]

ko

N/A

Thai

th

N/A

Arabic^[5]

ar

N/A

Basque

eu

N/A

Lithuanian

lt

N/A

Czech

cs

N/A

Indonesian

id

N/A

Polish

pl

N/A

Albanian

sq

N/A

Afrikaans

af

N/A

Belarusian

be

N/A

Bulgarian

bg

N/A

Burmese

my

N/A

Croatian

hr

N/A

Esperanto

eo

N/A

Estonian

et

N/A

Filipino

fil

N/A

Hebrew

he

N/A

Icelandic

is

N/A

Irish

ga

N/A

Kazakh

kk

N/A

Latvian

lv

N/A

Macedonian

mk

N/A

Malay

ms

N/A

Moldovan

ro

N/A

Mongolian

mn

N/A

Nepali^[5]

ne

N/A

Norwegian Bokmål^[4]

nb

Persian^[5]

fa

N/A

Serbian

sr

N/A

Slovak

sk

N/A

Slovenian

sl

N/A

Southern Sotho

st

N/A

Swahili

sw

N/A

Tagalog

tl

N/A

Tamil^[5]

ta

N/A

Ukrainian

uk

N/A

Uzbek

uz

N/A

Vietnamese

vi

N/A

Yiddish

yi

N/A

2. Wildcards in queries are not supported.

1. The index can also generate item summaries.

3. A specialized tokenizer based on dictionaries is used to split CJK characters into words, which can impact search results relevancy.

4. Will be detected as Norwegian (no).

5. Diacritics normalization isn’t supported. Stemming only matches terms with the same diacritics.

Troubleshooting language indexing issues

Note

Context and symptoms

You see inconsistent results when searching in a specific language on your website.

Cause

Most languages are limited to the Snowball stemming algorithms. Some rules are incomplete, meaning variations of words aren’t always stemmed correctly.

For example, when searching in German, the words passieren, passierend, and passiert might result in different stems, while passieren and passiere share the same stem.
The correct language isn’t being detected. Stemming is only applied if the language is detected for both the query and the indexed documents.

Resolution

Force the query-time language for the index by specifying the locale query parameter using an appropriate locale tag. For example, setting the locale parameter to de-DE forces the index to prioritize documents in German (as spoken in Germany) when returning results. For the index to take the locale into account, you must also set the forwardLanguageToCoveoIndex query parameter to true.
For Coveo ML CPR, RGA, SE and smart snippets models, see Multilingual content retrieval to configure your model to support multiple languages.