Vivoice Relabeled: Vietnamese Speech Recognition Data with Qwen3-ASR

Vivoice Relabeled is a speech dataset derived from the original capleaf/viVoice collection. The dataset has been processed using the Qwen/Qwen3-ASR-1.7B model to update audio-text labels, retaining samples with a Word Error Rate below 15%. It was uploaded by author JayLL13 to Hugging Face in March 2026.

Use Cases

• Training Vietnamese speech recognition models based on the described audio-text pairs.
• Benchmarking ASR model performance based on the relabeled audio samples.
• Fine-tuning or evaluating the Qwen3-ASR architecture based on the described processing method.
• Studying the impact of label refinement on WER metrics based on the described filtering criterion.

Strengths

• Audio samples have a sample rate of 44.1 kHz.
• Labels were updated using a specific ASR model (Qwen3-ASR-1.7B).
• Dataset includes a quality filter, retaining only samples with a Word Error Rate (WER) below 15%.

Limitations

• Description metadata is limited; actual data quality requires manual inspection after download.
• Column-level documentation is absent; field semantics must be inferred after download.
• Row count, file formats, and license information are unknown, which may limit suitability assessment.

Provenance

Source

Derived from the original dataset: capleaf/viVoice.

Collection Method

Relabeled using the Qwen/Qwen3-ASR-1.7B model, filtering samples with WER < 15%.

Time Range

null

Freshness

Last updated 2026-03-21 15:34:05; freshness should be verified.

Geography

null