Coursera Flash Sale
40% Off Coursera Plus for 3 Months!
Grab it
Explore discriminative language modeling techniques for Large Vocabulary Continuous Speech Recognition (LVCSR) in this 51-minute lecture by Murat Saraclar from the Center for Language & Speech Processing at Johns Hopkins University. Delve into advanced approaches that move beyond traditional generative language models by focusing on discriminative training methods that directly optimize recognition performance. Learn how these techniques can improve speech recognition accuracy by better modeling the decision boundaries between competing word sequences. Examine the theoretical foundations of discriminative language modeling, understand the differences from conventional n-gram models, and discover practical implementation strategies for large vocabulary speech recognition systems. Gain insights into the challenges and benefits of applying discriminative training to language modeling components within automatic speech recognition frameworks.
