You will listen to French audio clips, usually one to two minutes long, and check existing word-level and phoneme-level transcriptions against the audio. You will not transcribe audio from scratch.
Your work will help represent reading errors, pronunciation variations, and speech disfluencies accurately in datasets used to develop speech recognition and language-learning AI models.
- Validate existing ARPABET phonemic transcriptions against French audio.
- Check word-level and phoneme-level transcription accuracy.
- Identify and correct discrepancies in the annotations.
- Accurately mark reading errors, pronunciation variations, and speech disfluencies.
- Work through ongoing batches in a web-based annotation platform.
- Complete annotations that are reviewed through a peer quality assurance process.
What It Pays And Takes
- Pay: $0.80 per approved task.
- Higher accuracy and productivity can provide greater earning potential.
- Hours: At least 20 hours per week.
- Work type: Part-time contractor.
- Location: Open to candidates in France and Belgium.
- Language: Professional fluency in French.
- Skill level: Intermediate to expert proficiency in ARPABET phonemic transcription.
- Preferred background: Phonetics, linguistics, automatic speech recognition datasets, or child-speech transcription.
- Availability: Ready to start within one week.
- Requirement: Pass a qualification assessment.
How It Works
Apply on OpenTrain. The employer reviews applications there.
About AI Training Work
OpenTrain AI is the hiring and contracting organization for this role. AI training work uses examples reviewed and prepared by people to improve systems such as speech recognition models, and people with relevant language or technical expertise are paid for accurate, careful work.