Audiobook QA Expert — Dutch

What you would do

• Listen actively to AI-synthesized Dutch audiobooks assessing overall naturalness, phrasing, and intonation from a reader's perspective
• Identify and annotate specific errors: skipped or added words, mispronounced terms, incorrect numbers, abbreviation handling, unnatural prosody, audio glitches
• Categorize each flagged issue by error type and mark the precise timestamp or span in audio for model developers
• Report on listener experience: where does narration feel natural, where does it break engagement, how engaging is overall performance?
• Maintain accuracy and focus across many hours of continuous listening without fatigue

Who they want

• Native or near-native fluency in Dutch combined with genuine enthusiasm for reading and listening to audiobooks regularly
• Proficiency in working English sufficient to document errors, report findings, and communicate with team
• Prior experience with detail-oriented work: annotation, proofreading, transcription, or audio QA
• Ability to listen analytically while staying engaged catching subtle errors without losing narrative immersion
• Steady commitment to roughly 20 hours weekly with flexible scheduling

What the interview asks about

1. 1.Detecting subtle synthesis errors

   Some TTS errors are obvious while others are subtle - wrong stress patterns, misplaced pauses, or mispronounced proper nouns require linguistic knowledge and careful listening.

   For example: “AI narrates a compound word with unusual pauses between components. How do you categorize this error and what would you note?”

2. 2.Abbreviation and number handling

   Dutch has specific conventions for abbreviations, cardinals, and ordinals - TTS systems often get these wrong in language-specific ways.

   For example: “AI reads Dr. as Doctor then Doctorandus, and 25 augustus as twee vijf instead of vijfentwintig. How do you categorize these errors?”

3. 3.Sustaining focus over hours

   Audiobook evaluation requires listening 4+ hours daily - fatigue causes missed errors and inconsistency across sessions degrades quality.

   For example: “After 3 hours listening you notice errors in prosody you missed initially. How would you track your attention level and prevent fatigue degrading accuracy?”

4. 4.Listener experience vs. technical correctness

   An error is not just pronunciation correctness but whether it disrupts engagement with the story and cultural listener expectations.

   For example: “A name is pronounced correctly by one convention but unusually for Dutch speakers. The narration is technically right but feels off. How do you report this?”

5. 5.Proper nouns and foreign words

   Dutch audiobooks include place names, personal names, foreign words - getting these wrong especially damages listener experience and reveals model knowledge gaps.

   For example: “An audiobook contains French phrases and German place names pronounced with Dutch accent instead of natural cross-language handling. How do you flag this?”

A task you may get

Listen to a 30-minute Dutch audiobook sample and note all synthesis errors with type and location; write brief summary assessing naturalness and where narration works well or breaks engagement.

How to prepare

• Listen to a recent Dutch audiobook critically noting where narration could improve, where accents or phrasing felt off.
• Research common TTS error patterns: abbreviations, numbers, acronyms, proper nouns in Dutch and typical handling approaches.
• Familiarize yourself with annotation tools online; understand marking, categorizing, and documenting workflows.
• Read a transcribed audiobook chapter while listening; compare pronunciation of difficult words, abbreviations, and names to build reference for error detection.

The facts

Pay

$15–20/hr

Hours

Hourly, 20 hours a week

Where

Remote · Remote — preferred: Netherlands

Field

Language and Audio

Posted

9/14/2026

Places left

10

We wrote this page from the public Mercor listing. It may be out of date, so read the full posting before you apply.