Advanced level · Module 3: Open Models and Customisation · Lesson 4 of 4
AI for Local Languages
Assess AI support for local languages and plan responsible improvement.
In this lesson you will
- Measure model performance in local languages
- Plan data collection for language improvement ethically
Key concepts
language coverageevaluation in local languagescommunity dataconsent and ownership
The lesson
Most models perform much better in English and Arabic than in South Sudanese languages. Measure performance in the languages your users actually speak, with fluent reviewers.
Improving support needs text and speech data, collected with informed consent and respect for the communities who own the language.
Partner with language communities, universities and initiatives, agree how data will be used and shared, and return benefits to the communities.
Step by step
- Test a model on questions in two local languages.
- Have fluent speakers rate the answers.
- Identify the main problems.
- Draft an ethical data collection plan.
Practice activity
Write a local-language assessment for the capstone system.
Expected result: Measured results with fluent-speaker ratings and an ethical improvement plan.
Check your understanding
-
Who should evaluate answers in a local language?
Show answer
Fluent speakers.
-
What must language data collection respect?
Show answer
Informed consent and the communities' ownership and benefit.
Taught live by an instructor — online or in person. Free course lessons are for self-study and do not lead to a certificate — see registration options.