Krisp-AI/turn-taking-test-v1

Turn Taking Test Dataset

This dataset is collected for turn-taking benchmarking. The original recordings consisted of long-form conversations. These recordings were first reviewed and labeled manually by human annotators. After labeling, the conversations were segmented into shorter clips, ensuring that each segment preserved the context needed for accurate interpretation while maintaining alignment with the original annotations. The metadata contains the original filename, the last silence duration, speaker_id, age, gender, and the audio turn label: shift or hold.

The dataset language is English. Most speakers are U.S. English speakers, and there are also speakers with accented speech.

Audio files are single-channel, 16 kHz PCM (float32 format).

Key Statistics

Property	Value
Total audio duration	4 hours
Number of episodes	55
Number of speakers	30
Labeled "shift" cases	976
Labeled "hold" cases	1754

Quick Load

from datasets import load_dataset
from huggingface_hub import login

# Login with your token
login(token="your_hf_token")

ds = load_dataset("Krisp-AI/turn-taking-test-v1")
print(ds)
print(ds["test"][0])
print(ds["test"][0]["audio"])

License

See LICENSE file for the full license text.