NVVSpeech-Challenge/NVVSpeech-Challenge-Track1-Test-Set

NVVSpeech Challenge Track 1 Test Set

Track 1 test set for the NVVSpeech Challenge at ISCSLP 2026.

Task

Given a speech recording, produce a transcript that contains the spoken content and the non-verbal vocalization (NVV) tags at their corresponding positions.

Dataset Summary

Language	Samples
Chinese	985
English	961
Total	1,946

Files

.
├── README.md
├── SUBMISSION_GUIDE.txt
├── test.jsonl
├── ground_truth.jsonl
├── metadata.jsonl
└── audio/
    ├── zh_0001.wav
    ├── ...
    └── en_0961.wav

Each line in test.jsonl has the following form:

{"path": "audio/zh_0001.wav", "utt_id": "zh_0001", "text_with_nvvs": ""}

• path: relative path to the test audio.
• utt_id: unique utterance identifier; it must not be changed.
• text_with_nvvs: empty inference field to be filled by a participating system.

All 1,946 text_with_nvvs values in test.jsonl are intentionally empty. This file is the inference template distributed to participating systems.

ground_truth.jsonl contains the corresponding organizer-provided reference transcripts:

{"utt_id": "en_0001", "language": "en", "text_with_nvvs": "[burp] You see?"}

• utt_id: matches exactly one record in test.jsonl and one file in audio/.
• language: zh or en.
• text_with_nvvs: reference transcript with NVV tag markers.

The file contains one reference annotation for each test utterance.

metadata.jsonl links each recording to its language and reference transcript for the Hugging Face Dataset Viewer:

{"file_name": "audio/zh_0001.wav", "utt_id": "zh_0001", "language": "zh", "text_with_nvvs": "..."}

The file_name field is the relative path used by Hugging Face to display and play the corresponding audio file.

Canonical NVV Tags

breath, sniff, laugh, cry, cough, throat clearing, sneeze, sigh, gasp, snore, yawn, hum, moan, hiss, lipsmack, and burp.

Track 1 Evaluation

GitHub: Track 1 Evaluation Method