Provide detailed transcription and annotation of long-form podcast audio to enhance speech recognition system training and evaluation. This role requires precise speaker identification, timestamping, and quality assessment.
About this project
This project involves segmenting and transcribing podcasts typically lasting 30 to 60 minutes or more. The collected data supports the development of speech-to-text and voice recognition technologies across multiple languages and regions.
Responsibilities
- Transcribe podcast audio following specific project guidelines.
- Identify and label speakers consistently throughout the recordings.
- Segment speech and apply accurate timestamps.
- Note the perceived gender of speakers when required.
- Flag unclear audio, strong accents, incorrect languages, and synthetic or generated speech.
- Review and correct existing annotations as needed.
Requirements
- Native or near-native proficiency in the selected language and region.
- Excellent auditory attention to detail.
- Experience with transcription or speech annotation tasks.
- Access to a computer with reliable internet connectivity.
Project details
- Location: Remote work available for eligible languages and regions.
- Task duration: Podcasts generally range from 30 to 60 minutes or longer.
- Availability: Weekly task releases, assigned on a first-come, first-served basis; volume varies by location and batch.
Compensation
Payment is based on a fixed hourly rate determined by language and location.