Atyap Afwan_: Preserving Tyap Through Community-Driven Speech Data

Overview

This dataset contains audio recordings of the Tyap language, a Niger-Congo language spoken primarily in Nigeria. The dataset was collected by Raphael Musa Afana as part of the NaijaVoices Micro-Grants Heritage project.

Dataset Statistics

Total Recordings

• 98 audio recordings with corresponding transcript and translation files

• Audio files are organized by speaker in directories (TYAP1 through TYAP10)

• Each recording has a corresponding transcript file and translation file

Audio Duration

• Total duration: ~1 hour 9 minutes 47 seconds (4,186.66 seconds)

• Average duration per recording: ~42.3 seconds

• Duration range:

  • Minimum: ~7.7 seconds

  • Maximum: ~2 minutes 37 seconds (156.9 seconds)

Speakers

• 10 unique speakers contributing to the dataset

• Speaker distribution:

  • TYAP1: 16 recordings (16.3%)

  • TYAP2: 8 recordings (8.2%)

  • TYAP3: 9 recordings (9.2%)

  • TYAP4: 10 recordings (10.2%)

  • TYAP5: 14 recordings (14.3%)

  • TYAP6: 11 recordings (11.2%)

  • TYAP7: 5 recordings (5.1%)

  • TYAP8: 6 recordings (6.1%)

  • TYAP9: 10 recordings (10.2%)

  • TYAP10: 9 recordings (9.2%)

Gender Distribution

• Female speakers: 45 recordings (45.9%)

• Male speakers: 53 recordings (54.1%)

Age Range Distribution

• 18-29 years: 50 recordings (51.0%)

• 30-over years: 39 recordings (39.8%)

• Unspecified: 9 recordings (9.2%)

Geographic and Linguistic Information

• Country: All recordings are from Nigeria (98 recordings)

• Language: Tyap (98 recordings)

File Structure

The dataset is organized in the following directory structure:

├── audios/
│   ├── TYAP1/
│   │   ├── Tyap1_F_001.wav
│   │   ├── Tyap1_F_002.wav
│   │   └── ...
│   ├── TYAP2/
│   │   └── ...
│   └── TYAP{1-10}/
│       └── [audio files organized by speaker]
├── transcripts/
│   ├── TYAP1/
│   │   ├── Tyap1_F_001.txt
│   │   ├── Tyap1_F_002.txt
│   │   └── ...
│   ├── TYAP2/
│   │   └── ...
│   └── TYAP{1-10}/
│       └── [transcript files organized by speaker]
├── translations/
│   └── translation for Tyap language/
│       ├── TYAP1/
│       │   ├── Tyap1_F_001.txt
│       │   └── ...
│       └── TYAP{1-10}/
│           └── [translation files organized by speaker]
├── metadata.csv
└── dataset-card.md

Each recording consists of:

• Audio file (.wav format) organized in speaker-specific directories within audios/

• Transcript file (.txt format) containing the original Tyap transcription, organized in speaker-specific directories within transcripts/

• Translation file (.txt format) containing English translations, organized in speaker-specific directories within translations/translation for Tyap language/

• Metadata entry in metadata.csv with speaker information and file references

Text Encoding

All text files (transcripts, translations, and metadata CSV) are saved with UTF-8 with BOM (Byte Order Mark) encoding. This encoding choice is critical for preserving the linguistic integrity of the dataset:

• Diacritics and Non-ASCII Characters: The Tyap language uses diacritics and other non-ASCII characters that are essential for accurate representation of the language. UTF-8 with BOM ensures these characters are preserved without loss of detail.

• Software Compatibility: The BOM (Byte Order Mark) allows for proper character display when accessing files with Microsoft Excel and other software that may not automatically detect UTF-8 encoding. This ensures that users working with different tools can access the data without character corruption or loss.

When working with these files, it is recommended to use text editors and software that support UTF-8 with BOM encoding to maintain the full linguistic detail of the transcriptions and translations.

Metadata Fields

The metadata CSV file includes the following fields:

• SPEAKER ID: Unique identifier for each speaker (TYAP1 through TYAP10)

• GENDER: Gender of the speaker (Male/Female)

• AGE RANGE: Age range of the speaker (18-29, 30-over)

• COUNTRY: Country where recording was made (Nigeria)

• LANGUAGE: Language spoken (Tyap)

• AUDIO FILE NAME: Name of the audio file (e.g., Tyap1_F_001.wav)

• TRANSCRIPT: Name of the transcript file (e.g., Tyap1_F_001.txt)

• TRANSLATION: Name of the translation file (e.g., Tyap1_F_001.txt)

File Naming Convention

Files follow a consistent naming pattern:

• Format: {SpeakerID}_{Gender}_{SequenceNumber}.{extension}

• Example: Tyap1_F_001.wav represents the first recording from TYAP1 (Female speaker)

• Gender indicators: F for Female, M for Male

• Sequence numbers are zero-padded (001, 002, etc.)