README.md

---
language:
- en
- hi
- ar
- sw
- te
- tl
- pcm
- es
license: cc-by-4.0
gated: auto
extra_gated_heading: Access Yapdo-Mini
extra_gated_description: Please share your contact information to access this dataset.
  Access is granted immediately.
extra_gated_button_content: Agree and access dataset
extra_gated_fields:
  Affiliation:
    type: text
    required: true
  Use case:
    type: select
    options:
    - Research
    - Commercial
    - Education
    - label: Other
      value: other
task_categories:
- automatic-speech-recognition
- audio-classification
tags:
- conversational-speech
- multilingual
- accent
- code-switching
- multi-speaker
size_categories:
- n<1K
configs:
- config_name: default
  data_files:
  - split: train
    path: data/train-*
dataset_info:
  features:
  - name: audio
    dtype:
      audio:
        sampling_rate: 48000
  - name: text
    dtype: string
  - name: language
    dtype: string
  - name: accent
    dtype: string
  - name: relationship
    dtype: string
  - name: topics
    dtype: string
  - name: speech_characteristics
    dtype: string
  - name: num_speakers
    dtype: int64
  - name: duration_s
    dtype: float64
  - name: rms_dbfs
    dtype: float64
  - name: peak_amplitude
    dtype: float64
  - name: speech_ratio
    dtype: float64
  splits:
  - name: train
    num_bytes: 39764425.0
    num_examples: 12
  download_size: 30082744
  dataset_size: 39764425.0
---

# Yapdo-Mini

**Yapdo-Mini** is a sample of the [Yapdo](https://huggingface.co/datasets/liva-ai/yapdo) dataset, a conversational speech corpus drawn from 109,804 hours of recordings from 17,008 speakers across 67 languages. The source audio is natively recorded with separate speaker channels; the samples here are presented as combined conversations.

## Yapdo Data Highlights

| | |
|---|---|
| **Total audio** | 109,804 hours |
| **Unique speakers** | 17,008 |
| **Languages** | 67 |
| **Format** | 48 kHz, 16-bit PCM WAV per speaker |
| **Channel separation** | Each speaker on a dedicated, time-aligned track |
| **Speech type** | Spontaneous, unscripted, multi-party conversations |
| **Code-switching** | Yoruba-English, Hindi-English, Swahili-English ("Sheng"), Tagalog-Cebuano, and more |
| **Mean SNR** | ~33 dB |
| **Median RMS** | -26 dBFS |

### Top 10 Languages (estimated hours)

| Language | Hours | | Language | Hours |
|---|---|---|---|---|
| English | 31,660 | | Tagalog | 2,014 |
| Hindi | 8,412 | | Spanish | 1,651 |
| Arabic | 2,427 | | Nigerian Pidgin | 1,382 |
| Swahili | 2,075 | | Tamil | 1,288 |
| Hausa | 2,074 | | Cebuano | 848 |

*Note: These are estimated hours based on automated language detection. We are in the process of obtaining human-verified language and accent labels. The total number of languages and hours per language/accent are subject to change.*

<details>
<summary><strong>Hours by City (click to expand)</strong></summary>

Self-reported locations from speaker profiles across the full Yapdo dataset. Hours are total approved speaker-hours. '(unspecified)' means the user entered a country name rather than a specific city.

### Nigeria — 38,500 hours, ~7,917 speakers

| City | Hours | Speakers |
|---|---:|---:|
| Nigeria (unspecified) | 24,685.1 | 4,598 |
| Lagos | 5,564.9 | 1,355 |
| Abuja | 1,818.4 | 380 |
| Port Harcourt | 728.0 | 162 |
| Aba | 652.2 | 115 |
| Kaduna | 618.5 | 162 |
| Ibadan | 378.9 | 110 |
| Enugu | 365.2 | 77 |
| Benin | 352.7 | 50 |
| Kano | 266.2 | 70 |
| Uyo | 238.7 | 70 |
| Benin City | 224.5 | 64 |
| Warri | 206.8 | 51 |
| Ilorin | 194.3 | 44 |
| Minna | 194.2 | 49 |
| Jos | 186.0 | 48 |
| Bauchi | 179.0 | 21 |
| Owerri | 165.4 | 45 |
| Delta | 156.5 | 40 |
| Katsina | 156.1 | 37 |
| Kwara | 124.1 | 32 |
| Abia | 123.8 | 31 |
| Calabar | 110.3 | 49 |
| Abeokuta | 105.8 | 19 |
| Asaba | 102.7 | 12 |
| Ogun | 93.6 | 55 |
| Akure | 89.7 | 20 |
| Oyo | 60.4 | 24 |
| Yenagoa | 55.0 | 12 |
| Anambra | 53.6 | 28 |
| Ekiti | 42.3 | 13 |
| Ondo | 40.9 | 21 |
| Borno | 38.3 | 6 |
| Nasarawa | 34.1 | 5 |
| Osogbo | 27.9 | 12 |
| Maiduguri | 26.7 | 8 |
| Sokoto | 18.0 | 4 |
| Makurdi | 15.8 | 15 |
| Plateau | 5.4 | 3 |

### India — 15,608 hours, ~2,110 speakers

| City | Hours | Speakers |
|---|---:|---:|
| India (unspecified) | 11,899.1 | 1,370 |
| Delhi | 786.3 | 172 |
| Chennai | 579.1 | 44 |
| Mumbai | 349.4 | 77 |
| Hyderabad | 324.2 | 91 |
| Kolkata | 273.6 | 72 |
| Pune | 205.8 | 43 |
| Bangalore | 193.4 | 38 |
| Patna | 140.9 | 28 |
| Indore | 125.5 | 16 |
| Gurugram | 106.3 | 6 |
| Lucknow | 94.1 | 28 |
| Noida | 88.5 | 25 |
| Nagpur | 76.0 | 11 |
| Jaipur | 62.3 | 16 |
| Bhopal | 52.1 | 5 |
| Surat | 37.4 | 7 |
| Ranchi | 32.1 | 10 |
| Coimbatore | 26.6 | 3 |
| Varanasi | 25.7 | 2 |
| Kanpur | 23.7 | 7 |
| Chandigarh | 23.6 | 9 |
| Bengaluru | 22.6 | 9 |
| Ahmedabad | 21.4 | 7 |
| Visakhapatnam | 12.6 | 3 |
| Gwalior | 8.4 | 1 |
| Madurai | 6.0 | 3 |
| Kochi | 4.8 | 1 |
| Thiruvananthapuram | 3.9 | 2 |
| Mangalore | 2.4 | 4 |

### Philippines — 5,616 hours, ~664 speakers

| City | Hours | Speakers |
|---|---:|---:|
| Philippines (unspecified) | 4,492.0 | 460 |
| Manila | 415.9 | 60 |
| Davao | 301.3 | 70 |
| Cagayan De Oro | 133.6 | 18 |
| Cebu | 129.2 | 22 |
| Quezon City | 60.9 | 16 |
| Bulacan | 53.3 | 9 |
| Iloilo | 18.6 | 6 |
| Pampanga | 9.5 | 1 |
| Taguig | 2.1 | 2 |

### United States — 3,529 hours, ~531 speakers

| City | Hours | Speakers |
|---|---:|---:|
| United States (unspecified) | 2,676.8 | 383 |
| New York | 658.1 | 121 |
| California | 54.4 | 8 |
| Los Angeles | 49.8 | 6 |
| Florida | 37.8 | 1 |
| Miami | 22.9 | 3 |
| New Jersey | 18.8 | 2 |
| Virginia | 3.2 | 2 |
| Atlanta | 2.9 | 2 |
| Texas | 2.6 | 1 |
| Chicago | 2.1 | 2 |

### Indonesia — 3,110 hours, ~43 speakers

| City | Hours | Speakers |
|---|---:|---:|
| Indonesia (unspecified) | 3,107.3 | 41 |
| Jakarta | 2.9 | 2 |

### Kenya — 3,105 hours, ~483 speakers

| City | Hours | Speakers |
|---|---:|---:|
| Kenya (unspecified) | 2,136.4 | 226 |
| Nairobi | 888.9 | 230 |
| Mombasa | 41.4 | 11 |
| Nakuru | 28.3 | 6 |
| Eldoret | 8.1 | 8 |
| Kisumu | 1.6 | 2 |

### Egypt — 2,435 hours, ~293 speakers

| City | Hours | Speakers |
|---|---:|---:|
| Egypt (unspecified) | 1,985.8 | 192 |
| Cairo | 364.7 | 71 |
| Alexandria | 67.5 | 18 |
| Giza | 17.3 | 12 |

### Venezuela — 2,252 hours, ~92 speakers

| City | Hours | Speakers |
|---|---:|---:|
| Venezuela (unspecified) | 2,184.4 | 84 |
| Valencia | 62.1 | 3 |
| Caracas | 6.0 | 5 |

### Italy — 2,028 hours, ~30 speakers

| City | Hours | Speakers |
|---|---:|---:|
| Italy (unspecified) | 1,833.5 | 29 |
| Naples | 194.8 | 1 |

### Algeria — 504 hours, ~26 speakers

| City | Hours | Speakers |
|---|---:|---:|
| Algeria (unspecified) | 482.3 | 23 |
| Algiers | 21.4 | 3 |

### United Kingdom — 434 hours, ~55 speakers

| City | Hours | Speakers |
|---|---:|---:|
| United Kingdom (unspecified) | 293.2 | 38 |
| London | 135.9 | 14 |
| Birmingham | 4.9 | 3 |

### Pakistan — 327 hours, ~68 speakers

| City | Hours | Speakers |
|---|---:|---:|
| Pakistan (unspecified) | 267.3 | 46 |
| Lahore | 42.6 | 8 |
| Karachi | 13.8 | 11 |
| Faisalabad | 1.9 | 2 |
| Islamabad | 1.7 | 1 |

### Ghana — 300 hours, ~56 speakers

| City | Hours | Speakers |
|---|---:|---:|
| Ghana (unspecified) | 161.9 | 34 |
| Accra | 134.9 | 18 |
| Kumasi | 2.7 | 4 |

### South Africa — 200 hours, ~13 speakers

| City | Hours | Speakers |
|---|---:|---:|
| South Africa (unspecified) | 128.3 | 9 |
| Johannesburg | 36.1 | 2 |
| Cape Town | 32.2 | 1 |
| Pretoria | 2.9 | 1 |

### Bangladesh — 159 hours, ~22 speakers

| City | Hours | Speakers |
|---|---:|---:|
| Dhaka | 107.0 | 17 |
| Bangladesh (unspecified) | 52.1 | 5 |

### Colombia — 135 hours, ~13 speakers

| City | Hours | Speakers |
|---|---:|---:|
| Bogota | 112.3 | 8 |
| Colombia (unspecified) | 16.4 | 4 |
| Medellin | 5.9 | 1 |

### Other countries

| Country | Hours | Speakers |
|---|---:|---:|
| Malaysia | 99.3 | 3 |
| Mexico | 63.0 | 6 |
| Japan | 20.3 | 9 |
| Brazil | 16.3 | 2 |
| Cameroon | 3.3 | 3 |
| Morocco | 1.7 | 1 |

</details>

---

## Combined vs. Separated Audio

Each sample in this mini dataset is a **combined mix** of all speakers. The parent Yapdo corpus stores each speaker on a separate, time-aligned track. Here's what that difference sounds like — a Telugu conversation with 2 speakers:

### Combined (all speakers mixed)

<audio controls src="https://huggingface.co/datasets/liva-ai/yapdo-mini/resolve/main/audio_examples/combined_BXVavSaL0ieT.wav"></audio>

### Speaker 1 (isolated track)

<audio controls src="https://huggingface.co/datasets/liva-ai/yapdo-mini/resolve/main/audio_examples/separated_BXVavSaL0ieT_speaker1.wav"></audio>

### Speaker 2 (isolated track)

<audio controls src="https://huggingface.co/datasets/liva-ai/yapdo-mini/resolve/main/audio_examples/separated_BXVavSaL0ieT_speaker2.wav"></audio>

---

## All 12 Samples

| # | Language | Speakers | Variety | Transcript |
|---|---|---|---|---|
| 1 | sw | 2 | Nairobi urban | Yes |
| 2 | hi | 2 | | |
| 3 | tl | 2 | Central Visayas | |
| 4 | sw | 2 | Nairobi urban | Yes |
| 5 | ar | 3 | Cairene | |
| 6 | te | 2 | Karnataka/Bangalore | Yes |
| 7 | es | 3 | Venezuelan | |
| 8 | pcm | 2 | Nigerian English | Yes |
| 9 | en | 3 | Egyptian Arabic | Yes |
| 10 | pcm | 2 | Nigerian English | Yes |
| 11 | tl | 3 | Mindoreño | |
| 12 | en | 3 | Indian | |

---

## Schema

| Column | Type | Description |
|---|---|---|
| `audio` | `Audio(16kHz)` | Combined multi-speaker audio, 16 kHz mono |
| `text` | `string` | Timestamped transcript with speaker IDs (human-reviewed where available, otherwise empty). Full human-validated transcripts are available upon request. |
| `language` | `string` | Primary ISO 639-1 language code |
| `accent` | `string` | Accent or dialect label (e.g. "Nairobi urban", "Cairene", "Mindoreño") |
| `relationship` | `string` | Speaker relationship (friends, acquaintances, colleagues, etc.) |
| `topics` | `string` | Topics discussed |
| `speech_characteristics` | `string` | Notable audio features (code-switching, laughter, etc.) |
| `num_speakers` | `int` | Number of speakers in the clip |
| `duration_s` | `float` | Clip duration in seconds |
| `rms_dbfs` | `float` | RMS loudness in dBFS |
| `peak_amplitude` | `float` | Peak sample amplitude (0.0–1.0) |
| `speech_ratio` | `float` | Fraction of frames containing speech |

---

## Usage

```python
from datasets import load_dataset

ds = load_dataset("liva-ai/yapdo-mini", split="train")

for example in ds:
    print(f"{example['language']:>3s} | {example['num_speakers']} speakers | {example['accent']}")
    print(f"   Transcript: {example['text'][:100]}...")
    print()
```