ArashAzma's picture
Update README.md
e8a31b3 verified
|
raw
history blame contribute delete
1.46 kB
metadata
language:
  - fa
task_categories:
  - automatic-speech-recognition
  - text-to-speech
  - text-to-speech
dataset_info:
  features:
    - name: audio
      dtype: audio
    - name: transcription
      dtype: string
    - name: Length (seconds)
      dtype: int64
  splits:
    - name: train
      num_bytes: 368910411.993
      num_examples: 2101
  download_size: 365903240
  dataset_size: 368910411.993
configs:
  - config_name: default
    data_files:
      - split: train
        path: data/train-*
tags:
  - Benchmark
  - Speech
  - STT
  - ASR

This dataset consists of 3 hours of 16kHz audio collected from diverse environments to better represent real-world scenarios. The recordings were sourced from audiobooks, YouTube, and other public sources, ensuring a wide variety of speech styles and acoustic conditions.

One key advantage of this dataset is that it was collected from recent sources within the last few months, ensuring no overlap with training data and fairness for evaluating other STT models.

To enable a robust and fair comparison of models, the dataset has been carefully normalized: extra characters and inconsistencies in Persian text have been removed, and orthographic variations have been standardized.

Evaluation of different models on the dataset

For a detailed explanation of the normalization process, please refer to our GitHub page.