Words and the Company they keep
C(a,b) a b C(a,b) a b 11487 New York 80871 of the 7261 United Sates 58841 in th 5412 Los Angeles 26430 to the 3301 last year 21842 on the 3191 Saudi Arabia 21839 for the 2699 last week 18568 and the 16121 that the 2514 vice president 15630 at the 2378 Persian Gulf 15494 to be 2161 San Francisco 13899 in a 2106 Presiden Bush 13689 of a 2001 Middle East 13361 by the 1942 Saddan Hussein
Hypothesis Testing • Conclusions are about Ho • reject Ho in favour of H_1 • do no reject Ho • we never can conclude • reject H_1, or even • accept H_1
Results on an Hypothesis test Decision Reject Ho Don’t reject Ho Type I Error Righ Decision Truth H_1 Righ Decision Type II Error
Significance level • P(type I error) = significance level = α
Mutual Information and Word Classes
Markov Models
HMM algorithms: Trellis and Viterbi
Introduction to Parsing
Treebanks, Evaluation
PCFGs Introduction
Recommend
More recommend
