ETRI-Knowledge Sharing Plaform

Self-Improving Artificial Intelligence System Reduction Rate Automatic Speech Recognition(ASR) End-To-End Speech Recognition Error Reduction Training Data Recognition Model Speech Corpus Over Time Spontaneous Speech Korean Language Large Vocabulary Continuous Speech Recognition(LVCSR) Performance Levels Adaptive Parameter Deep Neural Network(DNN) Translation Model Speaker Adaptation Matrix Decomposition Target Domain Limited Data Online Services Open-Domain Dialog Text Normalization Data Scarcity Open Data Conventional Methods Large-Scale Performance Degradation Transformation Method Matching Pattern Machine Translation(MT) Korean Speech And Phrases Lm Adaptation Model Performance Model-Based New Method Post-Processing Code-Switching Clean Environment Computationally Expensive Translation Errors Deep Learning(DL) Language Model Adaptation Recognition Performance Singular Value Decomposition(SVD) Correction Rate Affine Transformation Automatic Conversion Linear Transformation Automatic Translation Need For