---
language: vi
tags:
- audio
- speech
- features
- wav2vec2
- representation
- tokenized
- alignment
- kmeans
license: mit
dataset_info:
  features:
  - name: domain
    dtype: string
  - name: text
    dtype: string
  - name: tokens
    sequence: int64
  splits:
  - name: wavjepa
    num_bytes: 1165942742.0
    num_examples: 285032
  download_size: 225850764
  dataset_size: 1165942742.0
configs:
- config_name: default
  data_files:
  - split: wavjepa
    path: data/wavjepa-*
---

# Vietnamese Wav2Vec2 Feature & K-Means Tokenized Dataset

This repository contains the structured speech features and tokenized cluster indices for the target `pw733` and clean `viVoice` Vietnamese datasets, formatted as Parquet tables.

## 📊 Dataset Schema
- **`audio_uuid`** (string): Unique identifier of the audio file.
- **`text`** (string): Transcription text (empty for raw `pw733` audio).
- **`features`** (list of list of float): Frame-level Wav2Vec2 embeddings (`[Num_Frames, 768]`).
- **`indices`** (list of int32): The mapped discrete token IDs matching the K-Means centroids (`[Num_Frames]`).
- **`dataset_origin`** (string): Source dataset name (`pw733` or `viVoice`).

## 🔗 Mapped K-Means Model
This dataset is tokenized using the 16,384 K-Means centroids hosted at:
👉 **[giangndm/wav2vec2-vietnamese-kmeans-16384](https://huggingface.co/giangndm/wav2vec2-vietnamese-kmeans-16384)**

## 🚀 Usage
```python
from datasets import load_dataset
ds = load_dataset("giangndm/audio-confidence-alignment")
print(ds["train"][0])
```

---
*Developed as part of the Multi-Project Voice Research Spike under `rd-lumi/luvox-vibe-research`.*