Transcribe 1–2 minute French audio clips into word- and phoneme-level ARPABET annotations for speech recognition and language-learning models. This part-time contract pays $18–$20 per hour and requires 20+ hours each week.
The work
You will turn batches of French audio clips into accurate word-level alignments and phoneme-level transcripts. Clips are usually 1–2 minutes long and are handled in a web-based annotation tool.
Your annotations will support speech recognition and language-learning models. Work is delivered in rolling batches and reviewed through a peer quality-assurance process before final acceptance.
- Produce or validate French ARPABET phonemic transcriptions.
- Identify reading errors, including misread words, substitutions, omissions, and insertions.
- Identify disfluencies such as filled pauses, repetitions, false starts, and self-corrections.
- Create consistent word-level alignments and phoneme-level transcripts.
- Follow peer quality-assurance feedback and revise work when needed.
What it pays and takes
This is a part-time contractor role paid for approved tasks. Higher throughput and accuracy can lead to higher earnings within the published rate.
- Pay: $18–$20 USD per hour.
- Time: 20+ hours per week on an ongoing basis.
- Location: Belgium or France, as specified for this project.
- Language: Native-level French and conversational English.
- Experience: Intermediate-level experience is listed; intermediate to expert ARPABET transcription ability is required.
- Start: You must be able to begin within one week of acceptance.
- Assessment: You must pass a qualification quiz before receiving production batches.
- Background: Experience in phonetics, linguistics, ASR datasets, phoneme-level labeling, or child-speech transcription is strongly preferred.
- Tools: You should be comfortable using a web-based annotation tool and following peer quality-assurance feedback.
How it works
Apply on OpenTrain. The employer reviews applications there.
About AI training work
AI training uses human-prepared examples to help software understand speech, text, images, and other data. In this role, your careful phoneme and speech annotations help improve how language models and speech recognition systems handle French audio.