A Hierarchical Bayesian Language Model based on Pitman-Yor Processes
Add to your list(s)
Download to your calendar using vCal
If you have a question about this talk, please contact Shakir Mohamed.
I will be discussing:
N-gram language modelling traditionally uses some form of “smoothing” technique to allocate some probability mass to unseen N-grams. Over the years people have come up with smoothing schemes that perform pretty well, but it’s not easy to get a handle on what they’re doing, and how to improve them.
In this paper, Teh shows that a hierarchical Bayesian language model with a very simplistic model of context performs pretty much as well as the current state of the art smoothing schemes, and in fact has strong similarities to an existing smoothing scheme.
This talk is part of the Machine Learning Reading Group @ CUED series.
This talk is included in these lists:
Note that ex-directory lists are not shown.
|