Coursera Flash Sale
40% Off Coursera Plus for 3 Months!
Grab it
Explore a cutting-edge approach to language modeling in this 48-minute lecture by Mari Ostendorf from the University of Washington. Delve into the concept of multi-factor context-aware language modeling, which addresses the challenge of language variation across different contexts. Learn about a novel mechanism for incorporating contextual factors into recurrent neural network (RNN) language models, allowing for efficient handling of both continuous and discrete variables. Discover how this approach improves upon existing methods by using context vectors to control low-rank transformations of the recurrent layer weight matrix. Examine experimental results demonstrating performance gains across various contexts and datasets, and understand the computational efficiency of this method compared to alternatives. Gain insights from Ostendorf's extensive research in dynamic statistical models for speech and language processing, and her work on integrating acoustic, prosodic, and language cues for speech understanding and generation.
