We are collecting and transcribing unscripted telephonic conversations across several regions to improve call-centre and customer service AI domains. This involves capturing natural speech, segmenting audio into manageable chunks, and providing high-accuracy verbatim transcriptions.

Open Opportunities by Region

We are currently seeking skilled speakers and transcriptionists in the following locations:

• Turkey (Turkish) – 330 Audio Hours
• Portugal (Portuguese-EU) – 320 Audio Hours
• Greece (Greek) – 320 Audio Hours

Key Responsibilities

• Data Collection: Capture natural, unscripted conversations with a diverse range of speakers (Ages 18-60).
• Segmentation: Timestamp audio files into segments (max 15 seconds) following natural sentence boundaries.
• Verbatim Transcription: Transcribe speech exactly as spoken, including fillers and disfluencies, maintaining a Word Error Rate (WER) of less than 5%
• Metadata Management: Ensure all data is delivered in a structured JSON format with mandatory fields like speaker ID and role.

What We’re Looking For

• Accuracy First: We strive for 95% Word Accuracy and 90% Tag Accuracy.
• Diversity: We aim for a 50/50 gender balance across all recordings to ensure our models are inclusive.
• Technical Savvy: Comfort working with WAV formats and following specific transcription conventions (e.g., using tags like [laugh], [music], or #uh for fillers).
• Reliability: The ability to complete reworks within a 48-hour window when quality adjustments are needed.

Why Join This Project?

This is a chance to work on a sophisticated AI development pilot. You’ll be part of a workflow that values authentic human speech and high-level linguistic precision. Payment is processed based on approved audio hours, ensuring a fair and transparent system for all contributors.

Apply – [email protected]