Multilingual AI Agent Behavior

Back to App Settings In the case of multi-lingual AI Agent, while much of the AI Agent behavior doesn’t change with the language, language semantics does impose some restrictions. This document lists the different behaviors that are language-dependent, so you as app developer can cater for such scenarios.

Pattern Support

Patterns can be used for Intent detection and Entity extraction. Those rules are for AI Agents in the English language. Though the multilingual AI Agents honor those rules, the language semantics impose some restrictions. Patterns can also be used in Small Talk (while defining queries), Knowledge Graph (using alternate question field), and Trait identification but their behavior might differ slightly.

Pattern syntax	Intent detection and entity extraction - Supported languages	Intent detection and entity extraction - Unsupported languages	Small talk	Traits	Knowledge graph
`word1 word2 … wordn` - This mandates all the words defined to be available in the user utterance in the same consecutive order with additional words allowed in between, before and after	All languages	-	Yes	Yes	Yes
`word1_word2` - Enforce phrase, no additional words allowed in between word1 and word2.	All languages	-	Yes	Yes	Yes
`word1 * word2` - 0 to infinite number of additional words between the specified words/phrases	All languages	-	Yes	Yes	Yes
`word1 *n word2` - Exactly n number of additional words between the specified words/phrases	All languages	-	Yes	Yes	Yes
`word1 *0 word2` - To disable wildcards between two tokens.	All languages	-	Yes	Yes	Yes
`word1 < word2` - Indicates the match for word2 should start from the beginning of a sentence.	All languages	-	Yes	Yes	Yes
`word1 > word2` - Indicates the end of the sentence and no words are allowed after it.	All languages	-	Yes	Yes	Yes
`!abc` - Indicates the word/concept “abc” shouldn’t exist anywhere in the user utterance after this token	All languages	-	Yes	Yes	Yes
`!!abc` - The very next word/concept shouldn’t be “abc”	All languages	-	Yes	Yes	Yes
`[...]` - Used to define a group of words/concepts and the match should be against exactly one of the group declared in `[ ]`.	All languages	-	Yes	Yes	Yes
`{...}` - Used to define an optional group or words/concepts and the match would be against zero or one of the words/patterns declared in `{ }`.	All languages	-	Yes	Yes	Yes
`(...)` - contain a pattern i.e when a pattern or part of a pattern is enclosed in these parentheses, we treat it as a pattern unlike `[ ]` and `{ }`.	All languages	-	Yes	Yes	Yes
`<<...>>` - Used to find words in any order	All languages	-	Yes	Yes	Yes
`word1` - If you quote words or use words that aren’t in canonical form, the system restricts itself to what you used in the pattern	All languages	-	No	No	No
System Concepts - Support for built-in concepts	English, German, Spanish, French	Portuguese (Brazilian), Italian, Chinese simplified, Chinese traditional, Indonesian, Korean, Dutch, Japanese, Arabic, Finnish, Russian, Polish, Ukrainian, Swedish, Kazakh, Norwegian, Hindi, Telugu, Tamil, Marathi, Catalan, Tagalog, Portuguese (European)	No	No	No
Developer concepts - Support for developer-defined concepts	English, German, Spanish, French, Portuguese (Brazilian), Italian, Chinese simplified, Chinese traditional, Indonesian, Korean, Dutch, Japanese, Arabic, Finnish, Russian, Polish, Ukrainian, Swedish, Kazakh, Norwegian, Hindi, Telugu, Tamil, Marathi, Catalan*	-	Yes	Yes	Yes*
Two-letter identifiers for Dates - Like `Mo` for Monday and `Tu` for Tuesday.	English, German, Spanish, Portuguese (Brazilian), Italian, Chinese simplified, Chinese traditional, Indonesian, Korean, Dutch, Japanese, Arabic, Finnish, Russian, Polish, Ukrainian, Swedish, Kazakh, Norwegian, Hindi	French, Telugu, Tamil, Marathi, Catalan, Tagalog, Portuguese (European)	-	-	No

* - In these cases, developer concepts aren’t supported in entity extraction

NLP Support

Here we will list out the NLP features and their support.

The Spell Correction feature isn’t supported in Universal Apps.

NLP feature	Supported languages	Unsupported languages	Small talk	Traits	Knowledge graph
List of Values
Full Match	All languages	-	-	-	No
Partial Match	All languages	-	-	-	No
Exact Match	All languages	-	-	-	No
Ambiguity behavior	All languages	-	-	-	No
Ambiguity with multi-item selection	English	Not supported in any language except English	-	-	No
Spell Correction	English, German, Spanish, French, Portuguese (Brazilian), Portuguese (European), Italian, Indonesian, Korean (from ver7.3), Dutch, Arabic, Finnish, Russian, Polish, Ukrainian, Swedish, Kazakh, Norwegian	Chinese simplified, Chinese traditional, Japanese, Hindi, Telugu, Tamil, Marathi	-	-	Yes
Amend Entities	English	Not supported in any language except English	-	-	Yes
Sentiment Analysis	English, German, Spanish, French, Dutch, Korean (from ver7.3), Japanese (from ver8.0), Russian (from ver8.0), Kazakh (from ver8.0), Polish (from ver9.2), Norwegian, Portuguese (European)	Portuguese (Brazilian), Italian, Chinese simplified, Chinese traditional, Indonesian, Arabic, Finnish, Ukrainian, Swedish, Hindi, Telugu, Tamil, Marathi, Catalan	-	-	Yes
App Synonyms (default)	English, German, Spanish, French	Portuguese (Brazilian), Portuguese (European), Italian, Chinese simplified, Chinese traditional, Indonesian, Korean, Dutch, Japanese, Arabic, Finnish, Russian, Polish, Ukrainian, Swedish, Kazakh	-	-	Yes
Knowledge Graph
Patterns	All Languages	-	-	-	Yes
Search in Answer	English, German (from ver7.3), Spanish (from ver7.3), French (from ver7.3), Portuguese (Brazilian) (from ver7.3), Italian (from ver7.3), Indonesian (from ver7.3), Dutch (from ver7.3), Finnish (from ver7.3), Polish (from ver7.3), Swedish (from ver7.3)	Chinese simplified, Chinese traditional, Korean, Japanese, Arabic, Russian, Ukrainian, Kazakh, Portuguese (European)	-	-	Yes
Spell Correction	English, German, Spanish, French, Portuguese (Brazilian), Portuguese (European), Italian, Indonesian, Korean, Dutch, Arabic, Finnish, Polish, Swedish, Ukrainian, Russian (from ver8.0), Kazakh (from ver8.0), Japanese (from ver8.0)	Chinese simplified, Chinese traditional, Catalan	-	-	Yes
Special Character support	All languages except English	English	-	-	Yes
NL Engines
Dependency Parser Model	German (from ver7.3), French (from ver7.3), Italian (from ver8.0)	English, Spanish, Portuguese (Brazilian), Chinese simplified, Chinese traditional, Indonesian, Korean, Dutch, Japanese, Arabic, Finnish, Russian, Polish, Ukrainian, Swedish, Kazakh, Norwegian, Hindi	-	-	-

​Pattern Support

​NLP Support

Pattern Support

NLP Support