README.md

---
language:
- id
tags:
- audio
- speech
- automatic-speech-recognition
- speech-recognition
task_categories:
- automatic-speech-recognition
pretty_name: Poseidon Indonesian Speech Dataset
size_categories:
- 1K<n<10K
dataset_info:
  features:
  - name: audio
    dtype: audio
  - name: file_id
    dtype: string
  - name: speaker_id
    dtype: string
  - name: language_code
    dtype: string
  - name: GT_transcript_english
    dtype: string
  - name: GT_transcript_native
    dtype: string
  - name: spoken_transcript_native
    dtype: string
  - name: spoken_transcript_english
    dtype: string
  - name: wer_score
    dtype: float64
  - name: cer_score
    dtype: float64
  - name: semantic_score
    dtype: float64
  - name: poseidon_score
    dtype: float64
  - name: duration
    dtype: float64
  - name: sampling_rate
    dtype: int64
  splits:
  - name: train
    num_bytes: 111012806059
    num_examples: 186536
  download_size: 111012806059
  dataset_size: 111012806059
configs:
- config_name: default
  data_files:
  - split: train
    path: data/train-*
---

# Poseidon Indonesian Speech Dataset

## Dataset Description

This dataset contains 186,536 Indonesian audio recordings from the Poseidon audio campaign.

### Dataset Statistics

- **Total Samples:** 186,536
- **Total Duration:** 1856 hours
- **Average WER:** 0.3714
- **Average CER:** 0.2510
- **Average Semantic Score:** 0.9594
- **Average Poseidon Score:** 0.8016

### Decision Rules

- **Poseidon score** (`poseidon_score`) **> 0.689** (*higher the better*)

### Language Distribution

- **id: 186,536 samples**

## Dataset Structure

### Data Fields

- `audio`: Audio file metadata and bytes
- `file_id`: Unique identifier for the audio file
- `speaker_id`: Unique identifier for the speaker
- `language_code`: ISO language code
- `GT_transcript_native`: Ground truth transcript in Indonesian
- `GT_transcript_english`: Ground truth transcript in English
- `spoken_transcript_native`: ASR-generated transcript in Indonesian
- `spoken_transcript_english`: ASR-generated transcript translated to English
- `wer_score`: Word Error Rate score
- `cer_score`: Character Error Rate score
- `semantic_score`: Semantic similarity score
- `poseidon_score`: Overall quality score
- `duration`: Audio duration in seconds
- `sampling_rate`: Audio sampling rate in Hz

### Data Splits

The dataset is delivered as a single `train` split (100% of the data).

## Usage

```python
from datasets import load_dataset

# Load the entire dataset
dataset = load_dataset("psdn-ai/psdn-voice-iiindonesian")

# Load specific split
train_data = load_dataset("psdn-ai/psdn-voice-iiindonesian", split="train")

# Access audio and metadata
sample = dataset["train"][0]
audio_array = sample["audio"]["array"]
sampling_rate = sample["audio"]["sampling_rate"]
transcript = sample["GT_transcript"]
duration = sample["duration"]
```

## Quality Metrics

This dataset bundles multiple quality indicators:

- **WER (Word Error Rate):** Measures word-level transcription accuracy
- **CER (Character Error Rate):** Measures character-level transcription accuracy
- **Semantic Score:** Measures semantic similarity between spoken and reference transcripts
- **Poseidon Score:** Composite quality score derived from the above metrics

## Filtering Examples

```python
from datasets import load_dataset

dataset = load_dataset("psdn-ai/psdn-voice-iiindonesian", split="train")

# Filter clips with low spam probability
human_sounding = dataset.filter(lambda x: x["poseidon_score"] > 0.689)
```

## Citation

```bibtex
@dataset{poseidon_indonesian_speech_dataset_2025,
  title={Poseidon Indonesian Speech Dataset},
  author={Poseidon-AI},
  year={2025},
  publisher={Poseidon-AI},
  howpublished={\url{https://huggingface.co/datasets/psdn-ai/psdn-voice-iiindonesian}}
}
```

## Contact

For questions or issues, please contact Poseidon team.