Update README.md
Browse files
README.md
CHANGED
|
@@ -134,3 +134,101 @@ configs:
|
|
| 134 |
- split: test
|
| 135 |
path: ms/test-*
|
| 136 |
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 134 |
- split: test
|
| 135 |
path: ms/test-*
|
| 136 |
---
|
| 137 |
+
|
| 138 |
+
# Semantic VAD / EoT — Sampled Test Set (`tchiayan/semantic-vad-eot`)
|
| 139 |
+
|
| 140 |
+
This dataset is a small, sampled derivative of [`Scicom-intl/semantic-vad-eot`](https://huggingface.co/datasets/Scicom-intl/semantic-vad-eot), created for quick, lightweight evaluation and benchmarking of Semantic VAD / end-of-turn (EoT) detection models without needing to stream or download the full source dataset.
|
| 141 |
+
|
| 142 |
+
Only the **`test`** split of the source dataset is used. For each subset (language / domain), **100 examples** are randomly sampled from a shuffle buffer, so this is intended for fast iteration and sanity-checking rather than as a full evaluation benchmark.
|
| 143 |
+
|
| 144 |
+
## Configs (subsets)
|
| 145 |
+
|
| 146 |
+
This repo provides three configs, each containing a single `test` split assembled from different combinations of the source dataset's subsets:
|
| 147 |
+
|
| 148 |
+
| Config | Included source subsets | Rows |
|
| 149 |
+
|---|---|---|
|
| 150 |
+
| `default` | `ms_dialects`, `ms_imda`, `ms_malaysian`, `ms_parliament` | 400 |
|
| 151 |
+
| `ms` | `ms_dialects`, `ms_imda`, `ms_malaysian`, `ms_parliament`, `ms_science_english` | 500 |
|
| 152 |
+
| `all` | `ms_dialects`, `ms_imda`, `ms_malaysian`, `ms_parliament`, `ms_science_english`, `en`, `fr`, `de`, `it`, `ja`, `ko`, `zh`, `pl`, `pt`, `ru`, `es`, `th`, `tr` | 1,800 |
|
| 153 |
+
|
| 154 |
+
`default` and `ms` focus on Malaysian speech (dialects, IMDA-derived, general Malaysian, parliamentary speech, and — for `ms`/`all` — Malaysian science/English content). `all` additionally covers the full set of languages available in the source dataset.
|
| 155 |
+
|
| 156 |
+
Load a specific config with:
|
| 157 |
+
|
| 158 |
+
```python
|
| 159 |
+
from datasets import load_dataset
|
| 160 |
+
|
| 161 |
+
ds = load_dataset("tchiayan/semantic-vad-eot", "ms", split="test")
|
| 162 |
+
```
|
| 163 |
+
|
| 164 |
+
## How this dataset was built
|
| 165 |
+
|
| 166 |
+
For every subset above, 100 rows were sampled from the source dataset's `test` split using streaming mode with a fixed shuffle buffer and seed, then tagged with their originating subset name (`ds_name`) and combined into a single `test` split per config. Audio was cast to 16 kHz.
|
| 167 |
+
|
| 168 |
+
```python
|
| 169 |
+
from datasets import load_dataset
|
| 170 |
+
|
| 171 |
+
split_names = [
|
| 172 |
+
"ms_dialects", "ms_imda", "ms_malaysian", "ms_parliament",
|
| 173 |
+
"ms_science_english", "en", "fr", "de", "it", "ja", "ko", "zh",
|
| 174 |
+
"pl", "pt", "ru", "es", "th", "tr",
|
| 175 |
+
]
|
| 176 |
+
|
| 177 |
+
ds_list = []
|
| 178 |
+
for split_name in split_names:
|
| 179 |
+
ds = (
|
| 180 |
+
load_dataset("Scicom-intl/semantic-vad-eot", split_name, split="test", streaming=True)
|
| 181 |
+
.shuffle(buffer_size=2500, seed=42)
|
| 182 |
+
.take(100)
|
| 183 |
+
.map(lambda x: {"ds_name": split_name})
|
| 184 |
+
)
|
| 185 |
+
ds_list.append(ds)
|
| 186 |
+
|
| 187 |
+
from datasets import Dataset, Audio
|
| 188 |
+
ds = Dataset.from_generator(lambda: (item for ds in ds_list for item in ds))
|
| 189 |
+
ds = ds.cast_column("audio", Audio(sampling_rate=16000))
|
| 190 |
+
```
|
| 191 |
+
|
| 192 |
+
The `default` and `ms` configs are simply subsets of the rows produced above, filtered to the relevant `ds_name` values, and `all` contains everything.
|
| 193 |
+
|
| 194 |
+
Because sampling uses a fixed seed (`42`) and buffer size (`2500`), the selection is reproducible against the source dataset, though results may shift if the upstream `Scicom-intl/semantic-vad-eot` dataset is updated.
|
| 195 |
+
|
| 196 |
+
## Dataset structure
|
| 197 |
+
|
| 198 |
+
Each row corresponds to a single audio segment together with its transcript, word-level alignment, and labeled silence spans that mark candidate end-of-turn / pause locations. This mirrors the schema of the source dataset with one additional column (`ds_name`).
|
| 199 |
+
|
| 200 |
+
| Column | Type | Description |
|
| 201 |
+
|---|---|---|
|
| 202 |
+
| `id` | `string` | Unique identifier for the example (source dataset ID). |
|
| 203 |
+
| `audio` | `audio` | Audio waveform, resampled to 16 kHz. |
|
| 204 |
+
| `language` | `string` | Language/locale code of the utterance (e.g. `en`, `ms`, `zh`). |
|
| 205 |
+
| `duration` | `float64` | Duration of the audio clip in seconds. |
|
| 206 |
+
| `silence_spans` | `list[{start, end}]` | Labeled silence intervals within the clip, in seconds — used as candidate EoT / pause points. |
|
| 207 |
+
| `words` | `list[{word, start, end}]` | Word-level forced-alignment timestamps for the transcript. |
|
| 208 |
+
| `messages` | `list[{role, content}]` | Conversational-style transcript (typically a single `user` turn) corresponding to the audio. |
|
| 209 |
+
| `ds_name` | `string` | Name of the originating subset from `Scicom-intl/semantic-vad-eot` (e.g. `ms_dialects`, `en`). |
|
| 210 |
+
|
| 211 |
+
## Intended use
|
| 212 |
+
|
| 213 |
+
This dataset is meant for:
|
| 214 |
+
- Quick local evaluation of Semantic VAD / end-of-turn detection models (e.g. against `livekit/eot-bench`-style pipelines) without pulling the full multi-million-row source dataset.
|
| 215 |
+
- Sanity-checking model behavior across Malaysian dialects/domains (`default`, `ms`) or a broader multilingual mix (`all`).
|
| 216 |
+
|
| 217 |
+
It is **not** a substitute for evaluating on the full `Scicom-intl/semantic-vad-eot` test split, given the small per-subset sample size (100 rows).
|
| 218 |
+
|
| 219 |
+
## Source dataset & license
|
| 220 |
+
|
| 221 |
+
This dataset is derived entirely from [`Scicom-intl/semantic-vad-eot`](https://huggingface.co/datasets/Scicom-intl/semantic-vad-eot) and inherits its **CC-BY-4.0** license. Please refer to the source dataset card for full details on data collection, forced alignment, and silence-span labeling methodology.
|
| 222 |
+
|
| 223 |
+
## Citation
|
| 224 |
+
|
| 225 |
+
If you use this dataset, please cite the original source dataset:
|
| 226 |
+
|
| 227 |
+
```
|
| 228 |
+
@misc{scicom_semantic_vad_eot,
|
| 229 |
+
title = {semantic-vad-eot},
|
| 230 |
+
author = {Scicom-intl},
|
| 231 |
+
year = {2025},
|
| 232 |
+
url = {https://huggingface.co/datasets/Scicom-intl/semantic-vad-eot}
|
| 233 |
+
}
|
| 234 |
+
```
|