Cantonese Language Evaluator

What you'll do

You work inside micro1's own evaluation tool rather than a customer platform, reviewing Cantonese text and speech for linguistic accuracy, fluency and whether it fits the context. When a model output is wrong or ambiguous, you say where and why, and you escalate the genuinely hard cases to the project lead.

Cantonese is a demanding language to evaluate: written Cantonese and the Standard Chinese most text models learn from diverge sharply, in vocabulary, particles and characters. A model can produce fluent-looking Chinese that no Hong Kong or Guangzhou speaker would say aloud. Catching that gap, in both text and speech, is the core of the job. The ad does not say which script (traditional or simplified) or which regional norm the project uses.

Around the evaluation work sit annotation, labelling and categorisation tasks for NLP development, ad hoc quality reviews as priorities shift, and notes on how the evaluation process itself could be improved.

Who fits

• Native Cantonese, written and spoken
• Experience in language evaluation, translation, linguistic annotation or language technology projects
• Exceptional written and verbal communication and close attention to detail
• An analytical approach that applies the same criteria to very different samples
• Comfort with online review tools and working independently across several assignments

Content moderation, transcription or quality assessment experience is a plus. The ad does not state an English level, though guidelines and feedback on a US platform are likely to be in English.

What it pays

$30–40/hour, micro1's published figure. It is a narrow band, which makes it more predictable than micro1's bilingual audio roles at $30–65, but the ceiling is lower. No weekly hours or project length are stated.

Worth knowing

Good:

• A tight, predictable band
• Mixed text and speech work, which is less monotonous than pure audio rating
• Cantonese specialists are scarce, so your skill is not interchangeable with Mandarin
• Remote, with no country restriction on the listing

Less good:

• Only eight openings
• The ad asks for prior evaluation or annotation experience, so this is less open than micro1's "no experience needed" bilingual roles
• Script and regional norm are unstated
• No hours, duration or start date published
• Contractor engagement: no benefits, your own tax to manage

About this listing

Posted by micro1 on its own jobs portal about two weeks before we confirmed it open on 24 September 2026. The $30–40/hour band, eight openings and the qualifications above are its published text. The customer is not named. See micro1.