Making Computers Good Listeners
Explore a lecture on advanced speech and language processing techniques presented by Joseph Keshet from TTI Chicago. Delve into novel approaches for improving machine listening and understanding capabilities, including a new theorem linking learning updates to performance measures. Examine applications in phoneme-to-speech alignment, phoneme recognition, and automatic voice onset time measurement. Learn how these techniques surpass traditional methods like Hidden Markov Models and approach human-level reliability in certain tasks. Discover potential implications for medical applications such as monitoring Parkinson's disease and predicting cancer treatment effectiveness. Gain insights into cutting-edge research at the intersection of machine learning and speech processing from an expert in the field.