jml2026/conversational-speech-dataset

---
license: cc-by-nc-4.0
language:
- en
task_categories:
- automatic-speech-recognition
- audio-classification
tags:
- voice-ai
- speech-data
- conversational-ai
- real-world-audio
- crowdsourced
- multi-speaker
- meeting-transcription
pretty_name: "🎙️ Silencio Conversational Speech Dataset"
configs:
- config_name: conversational_english
data_files:
- split: conversations
path: conversational_english/conversations/**
size_categories:
- n<1K
---

# 🎙️ Silencio Network: Conversational Speech Dataset

<p align="left">
<img src="https://cdn-uploads.huggingface.co/production/uploads/69162b50b89e7abe20de4b5a/LWhs4p2lPFcyiVsP0tluu.png" width="40%">
</p>

[](https://www.silencioai.com)
[](mailto:[e-mail])
[](mailto:[e-mail])

---

## Overview

Sample conversational speech data from Silencio Network's crowdsourced voice AI platform. This dataset contains **multi-speaker meeting recordings** with word-level transcripts, speaker diarization, and rich demographic metadata.

Each row represents one participant in a meeting and includes **3 audio files**:

| Audio Column | Description | Format |
|-------------|-------------|--------|
| `file_name` (speaker audio) | Individual participant's isolated recording | WAV |
| `full_meeting_single_channel` | Full meeting mixed to single channel | MP3 |
| `full_meeting_multi_channel` | Full meeting with separate speaker channels | WAV |

Plus **word-level meeting transcripts** with speaker turns and timestamps.

## Dataset Summary

| Config | Language | Meetings | Participants | Total Audio |
|--------|----------|----------|--------------|-------------|
| `conversational_english` | English | 4 | 8 | 135.9 MB |

## 🚀 Quick Start

```python
from datasets import load_dataset

# Load conversational English samples
ds = load_dataset("jml2026/conversational-speech-dataset", "conversational_english")

conversations = ds['conversations']

for sample in conversations:
speaker_audio = sample['audio'] # Individual speaker recording
meeting_audio = sample['full_meeting_single_channel'] # Full meeting
transcript = sample['meeting_transcript_text'] # Plain text transcript
print(f"Speaker {sample['speaker_id']} ({sample['gender']}, {sample['dialect']})")
print(f" Words spoken: {sample['speaker_word_count']}")
print(f" Meeting duration: {sample['meeting_duration']}s")
```

## Schema

| Column | Type | Description |
|--------|------|-------------|
| `file_name` | Audio | Individual speaker's isolated audio recording |
| `meeting_id` | int | Unique meeting identifier |
| `speaker_id` | string | Deterministic UUID for the speaker |
| `gender` | string | Speaker gender |
| `ethnicity` | string | Speaker ethnicity |
| `occupation` | string | Speaker occupation |
| `birth_place` | string | Speaker birth place |
| `dialect` | string | Speaker dialect |
| `year_of_birth` | int | Speaker year of birth |
| `years_at_birth_place` | int | Years lived at birth place |
| `languages_data` | string | JSON array of languages spoken with proficiency levels |
| `language` | string | Meeting language |
| `meeting_duration` | int | Meeting duration in seconds |
| `meeting_transcript_word_count` | int | Total word count of the meeting transcript |
| `speaker_word_count` | int | Word count for this speaker |
| `full_meeting_single_channel` | Audio | Full meeting audio mixed to single channel |
| `full_meeting_multi_channel` | Audio | Full meeting audio with separate speaker channels |
| `meeting_transcript_json` | string | Full meeting transcript as JSON with word-level timestamps and speaker IDs |
| `meeting_transcript_text` | string | Plain text meeting transcript with speaker turns |

## About Silencio Network

Silencio Network operates a global platform with **500,000+ contributors** across **130+ countries**, collecting voice data through a mobile app. The full corpus exceeds **100,000 hours** of validated speech data.

For access to the complete dataset or custom data collection, contact **[e-mail]**.

## License

This sample dataset is released under [CC BY-NC 4.0](https://creativecommons.org/licenses/by-nc/4.0/). Commercial licensing is available upon request.