KIT | KIT-Bibliothek | Impressum | Datenschutz

09: Kognitive Systeme, Vorlesung, SS 2017, 19.06.2017

Waibel, Alexander; KIT | Webcast [Hrsg.]

  • 0:00:00 Starten
  • 0:00:10 HMM Problems And Solutions
  • 0:02:34 HMMs In Speech Recognition
  • 0:04:42 Model Topologies
  • 0:05:49 Forward-Backward Training for Continuous Speech
  • 0:06:26 Discrete HHM's Vector Quantization
  • 0:08:37 Acoustic Modeling
  • 0:12:38 Neural Net Approaches to Pattern Classification
  • 0:12:47 Simple NN Vowel Classification
  • 0:13:20 HMM-DeepNN Hybrids
  • 0:14:24 Deep Neural Net Hybrids
  • 0:18:59 Time-Delay Neural Network (TDNN)
  • 0:26:51 Reverberation Robust Speech Reco
  • 0:27:08 TDNN / CNN - Waibel 1987
  • 0:28:53 Conversational Speech
  • 0:29:17 Convolutional Nets
  • 0:29:46 Convolutional Nets in Image Classification
  • 0:30:43 Mastering the Game of Go
  • 0:32:04 Speech Recognition (System Components)
  • 0:32:50 Dictionaries
  • 0:39:06 Language Models: Grammar Based
  • 0:40:39 Speech Recognition
  • 0:42:16 A Word Guessing Game
  • 0:43:00 Bigrams and Trigrams
  • 0:44:54 The Bag of Words Experiment
  • 0:45:10 Language Models: N-Grams
  • 0:46:56 Objective Estimation of Language Model Quality
  • 0:55:21 The Perplexity of a Language Model
  • 0:59:36 Recurrent Neural Nets
  • 1:00:25 Elman Networks - Simple RNN
  • 1:01:05 Jordan Networks - Simple RNN
  • 1:01:43 Backpropagation Through Time
  • 1:02:11 Modeling Sequences with RNN
  • 1:02:54 Measuring Recognizer Performance
  • 1:04:37 Factors Affecting Recognizer Performance
  • 1:04:49 How Good Does it Have to be?
  • 1:06:35 Voice Agents
  • 1:11:17 Natural Language Processing
  • 1:12:11 Machine Translation: Approaches
  • 1:15:17 Statistical Machine Translation
  • 1:18:47 RNN Encoder - Decoder
  • 1:19:42 Neural Machine Translation
  • 1:20:37 RNN Encoder-Decoder Architecture
  • 1:21:05 Attention Mechanism in the Recurrent Decoder
  • 1:21:28 BiRNN Encoder-Decoder with Attention

Zugehörige Institution(en) am KIT Institut für Anthropomatik und Robotik (IAR)
Publikationstyp Audio & Video
Publikationsdatum 20.06.2017
Erstellungsdatum 19.06.2017
Sprache Deutsch
DOI 10.5445/DIVA/2017-349
Identifikator KITopen-ID: 1000114748
Lizenz KITopen-Lizenz
Serie Kognitive Systeme, Vorlesung, SS 2017
Folge 9
KIT – Die Forschungsuniversität in der Helmholtz-Gemeinschaft
KITopen Landing Page