speech_recognition

Why we included this project

This is a compact look at how a Chinese speech recognizer is built from scratch, not a drop-in tool. It brings together TensorFlow, a CTC acoustic model, and the standard thchs30 Mandarin corpus, with simple train.py and test.py entry points and a config file you edit first. That makes it a useful walkthrough for developers and students who want to see how audio turns into Chinese text predictions without a framework hiding the details. The age is visible in the TensorFlow 1.5 and Python 3.5 stack, so plan to modernize APIs before running anything, but the core ideas are easy to follow. Treat it as a worked example of an early end-to-end Mandarin speech pipeline.