CS267
Chris Pollett
Nov 7, 2018
Let's see what happens as we vary the document length and `mu` in when doing Dirichlet smoothing...
Suppose the term "star" appears in document `d` of length 250 twice. We smooth the language model of `d` using the Corpus of Contemporary English in which star occurs 73695 times among 450,000,000 words.
Compute `M_d^mu(t)` for the case when (a) `mu=100`, (b) `mu=1000`, (c) `mu=10000`.
Suppose "star" did not appear in `d` and that `mu=1000` what would `M_d^mu(t)` be?
Post your answers to the Nov 7 Discussion Thread.