Turing is building an evaluation-grade dataset of multi-channel, multi-speaker Arabic conversations, and this role is the final QA pass on both the audio and the human-corrected transcripts. You audit channel isolation and recording fidelity, verify verbatim transcription and word-level timestamps in JSON, and issue pass/fail decisions with written justification.

Summary, eligibility guidance and screening preparation are prepared by Sidequest from the platform's posting. Check the source listing before applying.

The work

Contributors record unscripted group conversations in Arabic. Those recordings pass through an automatic transcription and diarization step, then through a human validation pass. You are the layer after that. For each session you listen to the multi-channel audio and judge whether it is technically usable — channel isolation with no bleed between speakers, a quiet recording environment, no clipping, no gain problems — and then read the transcript against the audio to confirm it is genuinely verbatim.

Verbatim here means unnormalized: stutters, fillers, repetitions, false starts, backchannels and abandoned clauses stay in. Prescriptive cleanup is an error, not a courtesy. You also check turn-level and word-level timestamp precision, speaker attribution through overlapping speech, and metadata fields — demographic markers, contextual domain, conversational tags. Sessions must be audited for PII, toxic or sensitive content. Every session ends in a pass/fail or agree/disagree status, and any disagreement with the annotator needs a comment specific enough for them to act on.

What the screen looks for

The platform is checking three things: native Arabic ability across dialects rather than Modern Standard Arabic alone; whether you can articulate what audio defects sound like and how you would spot them; and whether you are comfortable reading JSON-structured annotation output — segment objects, speaker labels, timestamp arrays — without needing a GUI to interpret it. Candidates from linguistics and phonetics, language teaching, and professional transcription are called out explicitly as strong fits.

Logistics

Fully remote, contractor engagement — no paid or medical leave.
Stated contract duration is 8 weeks, with possible extension based on performance and project need.
Pay band is not disclosed in the listing.
Shortlisted candidates are reviewed internally, then contacted for onboarding.
Expect headphone-based listening work with meaningful sustained-attention demands; the hours are asynchronous but the accuracy targets are not flexible.

Audio Transcribers - Audio Annotation & Diarization (Arabic)

The work

What the screen looks for

Logistics

Interview prep — 10 questions