Hugging Face's logo

Datasets:
mshojaei77
/
persian_tts_merged

Modalities:
Audio
Text
Formats:
parquet
Size:
10K - 100K
Libraries:
Datasets
Dask
Dataset card Viewer Files Community
1
persian_tts_merged / README.md
mshojaei77's picture
mshojaei77
Update README.md
2f94f8a verified
preview code
|
raw
history blame contribute delete
3.53 kB
metadata
dataset_info:
  features:
    - name: audio
      dtype:
        audio:
          sampling_rate: 16000
    - name: sentence
      dtype: string
  splits:
    - name: train
      num_bytes: 8012803500
      num_examples: 82215
  download_size: 7479090641
  dataset_size: 8012803500
configs:
  - config_name: default
    data_files:
      - split: train
        path: data/train-*

Merged Persian TTS Dataset

Overview

This dataset is a comprehensive collection of Persian speech data, merged from several high-quality sources to facilitate Text-to-Speech (TTS) research and development for the Persian language. It combines audio recordings with their corresponding transcriptions, providing a rich resource for training and evaluating Persian TTS systems.

Dataset Details

  • Language: Persian (Farsi)
  • Total Samples: [Insert total number of samples after merging]
  • Audio Format: WAV (16-bit PCM)
  • Sampling Rate: 16000 Hz
  • Columns:
    • audio: Audio file in WAV format
    • sentence: Corresponding transcription of the audio in Persian text

Source Datasets

This merged dataset combines data from the following sources:

  1. Persian Common Voice 17.0

  2. Persian Concatenated

  3. Persian Speech Dataset

  4. SP_HW5_PersianTTS

  5. General Persian Dataset

Preprocessing

The following preprocessing steps were applied to create this merged dataset:

  1. Audio files were standardized to a 16000 Hz sampling rate.
  2. Text transcriptions were unified under the column name 'sentence'.
  3. Only 'audio' and 'sentence' columns were retained from each source dataset.

Usage

This dataset can be easily loaded using the Hugging Face datasets library:

from datasets import load_dataset

dataset = load_dataset("mshojaei77/persian_tts_merged")

Applications

This dataset is suitable for various speech-related tasks in Persian, including but not limited to:

  • Text-to-Speech (TTS) system development
  • Speech recognition model training
  • Voice activity detection
  • Linguistic analysis of Persian speech patterns

Ethical Considerations

Users of this dataset should be aware of and respect the following ethical considerations:

  • Privacy: Ensure that any use of this dataset complies with privacy regulations.
  • Bias: Be aware of potential biases in the dataset, such as accent or dialect representation.
  • Responsible AI: Use this dataset to develop AI systems that respect and preserve Persian language and culture.

Acknowledgements

We express our gratitude to the creators and contributors of the original datasets, without whom this merged dataset would not have been possible.