Call Center Audio With 13,000 Hours Of Customer Service Calls

Call Center Audio is a large audio dataset containing over 13,000 hours of real-world customer service calls. It features time-stamped transcripts and over 90% unique speakers, supporting tasks like speech recognition and speaker diarization. The dataset was created by ud-nlp and was last updated in March 2026.

Use Cases

• Train automatic speech recognition models on 13,000+ hours of audio with time-stamped transcripts for accurate transcription.
• Develop speaker diarization systems leveraging the dataset's feature of over 90% unique speakers within customer service calls.
• Build conversational AI models using the real customer service call audio and associated transcripts for dialogue simulation.
• Fine-tune text-to-audio models for call center applications using the paired audio and transcript data.

Strengths

• Contains over 13,000 hours of audio data, providing substantial scale for model training.
• Features time-stamped transcripts, enabling precise alignment between audio and text for speech tasks.
• Includes over 90% unique speakers, offering high speaker diversity for diarization and recognition models.

Limitations

• The specific geographic origin of calls is unclear beyond a 'Regionus' tag, which may indicate a US bias.
• The license is listed as 'cc By Nc Nd 40', which restricts commercial use and derivative works.
• Sample data and detailed column or file format information are unavailable for preliminary inspection.

Provenance

Source

ud-nlp via Hugging Face.

Collection Method

Collection of real customer service calls from global call centers.

Freshness

Last updated on 2026-03-23.

Geography

Global call centers, with a tag indicating 'Regionus'.