Towards a Computational Model of Human Speech Recognition

Explore a comprehensive lecture on developing computational models for human speech recognition presented by Partha Niyogi from Bell Labs in 2000. Delve into the challenges of aligning computer-based speech recognition with human cognitive processes. Examine the potential of distinctive feature-based approaches and the integration of statistical learning models to bridge the gap between linguistic invariance and acoustic variability. Learn about phonetic-feature learning models within Vapnik's structural risk minimization framework and their applications to various speech problems. Gain insights from Niyogi's extensive background in EECS, learning theory, and speech and language research, including his work on language acquisition, formal learning theory, and speech perception. This 1 hour 32 minute talk, hosted by the Center for Language & Speech Processing at Johns Hopkins University, offers a deep dive into the intersection of human speech recognition and computational modeling.