mtasr 1.0.1.post1

MT ASR interface for developers

MT ASR demos for developers

This repository contains the implementation of a gRPC client to require the result of the streaming result from mthreads.

Supported languages:

• Python

Python Usage

Install dependencies

pip install mtasr

Run demos

The ASR examples are in egs/.

Note: You may need to change the URL and token in the demo script to access the ASR server.

If the authorization is required, please use your token in the examples.

In the examples, it read the audio in demo.wav and get the recognized result from ASR server.

You can replace the audio file or create your own applications.

1. Real-time streaming ASR

   python egs/realtime_asr_demo.py
   

   During sending audio data,

   During receiving ASR result, test_on_sentence_changed and test_on_sentence_end return the recognized result which is a dictionary.

   For test_on_sentence_changed, result only contains the recognized content:

    [{'sentence': '你好'}]
   

   For test_on_sentence_end, result contains the text and the word-level information:

    [{'sentence': '你好欢迎使用摩尔线程语音识别', 'wordpieces': [{'word': '你', 'start': 860, 'end': 960}, {'word': '好', 'start': 1060, 'end': 1160}, {'word': '欢', 'start': 1380, 'end': 1480}, {'word': '迎', 'start': 1500, 'end': 1600}, {'word': '使', 'start': 1660, 'end': 1760}, {'word': '用', 'start': 1820, 'end': 1920}, {'word': '摩', 'start': 2020, 'end': 2120}, {'word': '尔', 'start': 2180, 'end': 2280}, {'word': '线', 'start': 2460, 'end': 2560}, {'word': '程', 'start': 2620, 'end': 2720}, {'word': '语', 'start': 3100, 'end': 3240}, {'word': '音', 'start': 3240, 'end': 3280}, {'word': '识', 'start': 3340, 'end': 3440}, {'word': '别', 'start': 3500, 'end': 3600}]}]
   

2. One-sentence streaming ASR

   python egs/one_sentence_asr_demo.py
   

   the result is similar to that of the realtime ASR.