@@ -7,15 +7,15 @@ The minimum PaddlePaddle version needed for the code sample in this directory is
...
@@ -7,15 +7,15 @@ The minimum PaddlePaddle version needed for the code sample in this directory is
DeepASR (Deep Automatic Speech Recognition) is a speech recognition system based on PaddlePaddle FLuid and Kaldi. It uses the Fluid framework to perform the configuration and training of acoustic models in speech recognition and integrates Kaldi decoder. It is designed to facilitate Kaldi users to implement the rapid and large-scale training of acoustic models, and to use Kaldi to complete complex speech data preprocessing and final decoding processes.
DeepASR (Deep Automatic Speech Recognition) is a speech recognition system based on PaddlePaddle FLuid and Kaldi. It uses the Fluid framework to perform the configuration and training of acoustic models in speech recognition and integrates Kaldi decoder. It is designed to facilitate Kaldi users to implement the rapid and large-scale training of acoustic models, and to use Kaldi to complete complex speech data preprocessing and final decoding processes.
### Content
### Content
-[Introduction](#Introduction)
-[Introduction](#introduction)
-[Installation](#Installation)
-[Installation](#installation)
-[Data reprocessing](#Datareprocessing)
-[Data reprocessing](#data-reprocessing)
-[Training](#Training)
-[Training](#training)
-[Perf profiling](#Perfprofiling)
-[Perf profiling](#perf-profiling)
-[Inference & Decoding](#Inference & Decoding)
-[Inference and Decoding](#inference-and-decoding)
-[Scoring error rate](#Scoring error rate)
-[Scoring error rate](#scoring-error-rate)
-[Aishell example](#Aishellexample)
-[Aishell example](#aishell-example)
-[Question and Contribution](#Question and Contribution)
-[Question and Contribution](#question-and-contribution)
After fully training the acoustic model, using the model checkpoint saved in the training process, the input audio data can be decoded and output, and the sound to text recognition result can be obtained.
After fully training the acoustic model, using the model checkpoint saved in the training process, the input audio data can be decoded and output, and the sound to text recognition result can be obtained.
