ASR-Training-Data-Chunker

April 13, 2025 ยท View on GitHub

A Python tool for chunking text files into smaller segments based on speaking speed and desired duration, optimized for ASR (Automatic Speech Recognition) training data preparation.

Purpose

This tool helps you split large text files into smaller chunks that:

  • Take a specific amount of time to read aloud based on your speaking speed
  • Break at logical sentence boundaries to maintain context
  • Can be used for recording voice samples for ASR training

Installation

No installation required. Simply clone this repository:

git clone https://github.com/danielrosehill/ASR-Training-Data-Chunker.git
cd ASR-Training-Data-Chunker

Usage

  1. Place your source text file in the chunker/source/ directory as source.txt
  2. Run the chunker script:
python chunker.py

  1. Follow the prompts to enter:

    • Your speaking speed in words per minute (WPM)
    • Desired chunk duration in minutes

  2. The script will create chunked files in the chunker/chunked/ directory

Determining Your Speaking Speed

You can measure your speaking speed using online tools like:

Average speaking speeds:

  • Slow: ~120 WPM
  • Average: ~150 WPM
  • Fast: ~180+ WPM

Example

For a text with 9000 words:

  • Speaking speed: 150 WPM
  • Desired chunk duration: 2 minutes

The script will create approximately 30 chunks, each containing around 300 words and taking about 2 minutes to read aloud.

License

MIT