#Languages

TechWolf supports international organisations by working in multiple languages. Language is set in three independent places: the language of the input data TechWolf reads, the language skill names and descriptions are displayed in, and the language of the tools people work in. Because each is configured separately, an organisation can load data in one language and present skills to its employees in another.

#Key attributes

The table below shows where each language is available. The surface columns show the language of the interface itself. For skill names and descriptions, the tenant's vocabulary decides which languages can be asked for, and the reader picks one from that set, which is why they are listed per vocabulary.

#How it works

Input data. The language of each piece of input data is declared when the data is loaded, and it can also be detected automatically. TechWolf's skill extraction models are multilingual, and skills are inferred from the text in the language it was written in. What comes out of inference are vocabulary skills, so they are displayed in whichever language the reader has selected, whatever language the source data was in.

Skill names and descriptions. These live in the skill vocabulary, which carries a name and a description per language. Any consumer asks for a language per request, and UK English is used when none is asked for. In the Console, each user picks their own language for skill names and descriptions, from the languages the tenant's vocabulary carries. A tenant running the Workday vocabulary uses Workday's own translations: description coverage varies by language there, and where a skill has no description in the selected language, its English description is shown instead.

Console. The Console interface itself is in UK English.

Assistant. The Assistant messages each employee in the language they are configured for, falling back to English where a language is not set, and the languages an organisation offers are chosen per tenant.

Analyst Agent. The agent's interface is in English. The conversation itself is not limited to it: ask the agent to answer in another language and it will.

Evaluation tool. The evaluation itself is in English. The language for skill names and skill descriptions is chosen when an evaluation is set up, from the languages the tenant's vocabulary carries.

Work Intelligence. Task names and descriptions are held in a single language, UK English.

Programmatic access. Through the API, the display language is chosen per request, and the taxonomy endpoints list the languages available in the tenant's vocabulary, per vocabulary. Standard IETF language tags are accepted and mapped to their equivalents. Asking for a language the tenant's vocabulary does not carry returns an error rather than falling back to English; the English fallback applies only to individual skills that have no entry in a language the vocabulary does support.

#Prerequisites

Languages is part of Configurations. Input data carries its language per load, or has it detected automatically. Displaying a skill name or description in a language requires that language to be part of the tenant's skill vocabulary.

#Availability

• Console: a per-user language setting for skill names and descriptions; the interface itself is in UK English.

• Assistant: per-tenant message languages, per-employee delivery language, and per-tenant overrides of the assistant's own wording.

• Agent: the interface is in English, the agent answers in another language when asked, and skill names and descriptions come in the vocabulary's languages.

• Exports: skill names and descriptions are delivered in the tenant's vocabulary languages.

• API: a per-request display language, a per-load input language with automatic detection, and the list of languages available in the tenant's vocabulary.