Omarrran/kashmiri_words_dataset_restructured

Frequently Spoken Kashmiri Words Dataset

This dataset contains processed audio data of 12 frequently spoken words in Kashmiri, collected from native speakers across three major regions of Kashmir. The dataset was used in research for Kashmiri spoken word recognition using advanced techniques such as dual feature extraction (MFCC + partial Mel-Spectrogram) and a hybrid CNN-gMLP model. It includes train/test splits, audio files, and one-hot encoded labels.

Summary of the Dataset: File Count: 5 files Size: 369.73 MB Words: 12 Kashmiri words Train Samples: 1151 Test Samples: 169

About this directory Data Format

Restructured Dataset Information:

• Root Directory: Contains metadata files and split directories.

  • README.md: This file, providing an overview of the dataset.
  • dataset_info.json: Contains metadata about the dataset, including features and split information.
  • label_dict.json: Dictionary mapping label indices to the corresponding Kashmiri words.
  • dataset_summary.csv: A summary of the dataset, including statistics for each example.

• train: Contains the training data.

  • train.arrow: Arrow file containing preprocessed spectrogram data and labels for training.

• test: Contains the test data.

  • test.arrow: Arrow file containing preprocessed spectrogram data and labels for testing.

File Descriptions:

• train.arrow:

  • Contains preprocessed spectrogram data for training.
  • Includes corresponding labels (originally one-hot encoded).
  • Stored in Apache Arrow format for efficient data loading and processing.

• test.arrow:

  • Contains preprocessed spectrogram data for testing.
  • Includes corresponding labels (originally one-hot encoded).
  • Stored in Apache Arrow format for efficient data loading and processing.

• dataset_info.json:

  • Provides metadata about the dataset, including feature descriptions and split information.

• label_dict.json:

  • Dictionary mapping label indices to the corresponding Kashmiri words.
  • Useful for interpreting the numeric labels in the dataset.

• dataset_summary.csv:

  • Provides a summary of each example in the dataset.
  • Includes statistics such as data shape, mean, standard deviation, min, and max values for both data and labels.
  • Useful for getting an overview of the dataset characteristics.

"""

Citation

@dataset{kashmiri_words_dataset,
  author       = {[Haq Nawaz Malik]},
  title        = {Frequently Spoken Kashmiri Words Dataset},
  year         = {2024},
  publisher    = {Hugging Face},
  version      = {1.0.0},
  url          = {https://huggingface.co/datasets/Omarrran/kashmiri_words_dataset_restructured}
}

License

[MIT]