TR-Podcast-dataset

This is a merged speech dataset containing 115031 audio segments from 281 source datasets.

Dataset Information

• Total Segments: 115031
• Speakers: 825
• Languages: tr
• Emotions: happy, neutral, angry, sad
• Original Datasets: 281

Dataset Structure

Each example contains:

• audio: Audio file (WAV format, original sampling rate preserved)
• text: Transcription of the audio
• speaker_id: Unique speaker identifier (made unique across all merged datasets)
• language: Language code (en, es, fr, etc.)
• emotion: Detected emotion (neutral, happy, sad, etc.)
• original_dataset: Name of the source dataset this segment came from
• original_filename: Original filename in the source dataset
• start_time: Start time of the segment in seconds
• end_time: End time of the segment in seconds
• duration: Duration of the segment in seconds

Usage

Loading the Dataset

from datasets import load_dataset

# Load the dataset
dataset = load_dataset("Codyfederer/tr-podcast-dataset")

# Access the training split
train_data = dataset["train"]

# Example: Get first sample
sample = train_data[0]
print(f"Text: {sample['text']}")
print(f"Speaker: {sample['speaker_id']}")
print(f"Language: {sample['language']}")
print(f"Emotion: {sample['emotion']}")
print(f"Original Dataset: {sample['original_dataset']}")
print(f"Duration: {sample['duration']}s")

# Play audio (requires audio libraries)
# sample['audio']['array'] contains the audio data
# sample['audio']['sampling_rate'] contains the sampling rate

Alternative: Load from JSONL

from datasets import Dataset, Audio, Features, Value
import json

# Load the JSONL file
rows = []
with open("data.jsonl", "r", encoding="utf-8") as f:
    for line in f:
        rows.append(json.loads(line))

features = Features({
    "audio": Audio(sampling_rate=None),
    "text": Value("string"),
    "speaker_id": Value("string"),
    "language": Value("string"),
    "emotion": Value("string"),
    "original_dataset": Value("string"),
    "original_filename": Value("string"),
    "start_time": Value("float32"),
    "end_time": Value("float32"),
    "duration": Value("float32")
})

dataset = Dataset.from_list(rows, features=features)

Dataset Structure

The dataset includes:

• data.jsonl - Main dataset file with all columns (JSON Lines)
• *.wav - Audio files under audio_XXX/ subdirectories
• load_dataset.txt - Python script for loading the dataset (rename to .py to use)

JSONL keys:

• audio: Relative audio path (e.g., audio_000/segment_000000_speaker_0.wav)
• text: Transcription of the audio
• speaker_id: Unique speaker identifier
• language: Language code
• emotion: Detected emotion
• original_dataset: Name of the source dataset
• original_filename: Original filename in the source dataset
• start_time: Start time of the segment in seconds
• end_time: End time of the segment in seconds
• duration: Duration of the segment in seconds

Speaker ID Mapping

Speaker IDs have been made unique across all merged datasets to avoid conflicts. For example:

• Original Dataset A: speaker_0, speaker_1
• Original Dataset B: speaker_0, speaker_1
• Merged Dataset: speaker_0, speaker_1, speaker_2, speaker_3

Original dataset information is preserved in the metadata for reference.

Data Quality

This dataset was created using the Vyvo Dataset Builder with:

• Automatic transcription and diarization
• Quality filtering for audio segments
• Music and noise filtering
• Emotion detection
• Language identification

License

This dataset is released under the Creative Commons Attribution 4.0 International License (CC BY 4.0).

Citation

@dataset{vyvo_merged_dataset,
  title={TR-Podcast-dataset},
  author={Vyvo Dataset Builder},
  year={2025},
  url={https://huggingface.co/datasets/Codyfederer/tr-podcast-dataset}
}

This dataset was created using the Vyvo Dataset Builder tool.