French Audio Transcription Auditor for AI Training

Help improve French-language AI training data by evaluating the accuracy of speech transcriptions and word-level audio alignment. This is detailed listening and quality-review work, with every audit decision supported by a clear written rationale. Role Overview

You will audit two audio data collections:

• Multilingual Transcription Audit: listen to human or agent speech, compare it with an annotator-created transcript, determine whether it accurately reflects what was said, apply the required error codes, issue a pass or fail decision, and explain your reasoning.
• Forced Alignment Audit: review audio already segmented at the word level, confirm that each word''s start and end timestamps match the spoken audio, and verify that the reference transcription matches the spoken form. When needed, correct boundaries, split or merge segments, add missing segments, or remove segments that contain no word.

• Listen to audio, assess it against written standards, and submit complete audit decisions.
• Apply every relevant error code when an issue is found, rather than selecting only the first error.
• Write plain-language pass or fail rationales, identifying critical errors first so the basis for a failed review is clear.
• Complete a short calibration phase before beginning production work.

• Native or near-native fluency in French as spoken in France. You must have been born and raised in a region where this language variety is dominant, or be fully fluent with at least five years of residence in such a region.
• Strong English reading and writing skills to interpret detailed guidelines, error codes, and provide clear written feedback.
• Ability to apply a written rubric consistently across hundreds of judgments.
• Careful listening skills, including distinguishing similar sounds, identifying word boundaries in continuous speech, and separating genuine acoustic ambiguity from clear annotation errors.
• Experience in transcription, subtitling, localization, linguistic annotation, or quality assurance is highly relevant.
• Helpful but not required: AI data annotation or model evaluation experience, phonetics or linguistics knowledge, and familiarity with audio editing or waveform tools.

• Remote, hourly engagement.
• Each collection runs at 5.00 audit hours per task.

$41.50 per hour.