Quizz 2 - Parameters of the probabilistic model

Quizz 2 - Parameters of the probabilistic model

by Leon Christian Muller -
Number of replies: 1

Hello,

I have a question about the latest quizz. I guess I'm wrong on what I'll say since I failed the question, but I find the real answers differing from the slides.

I didn't understand why do we not have to take into parameters these two: P(wn1|Tn1) and P(Tn1), since they are literally computed from the ones that are the actual correct answers: P(Ti|Ti−1) for all 1<i≤n, P(wi|Ti) for all 1≤i≤n and P(T1). Couldn't they also be parameters?


Here is the slide in the course that makes me wonder (slide 14 of POS Tagging). The last equation, that is actually equal to P(Tn1|wn1) (since P(wn1) should be fixed), is basically the equality of both my answers.


In reply to Leon Christian Muller

Re: Quizz 2 - Parameters of the probabilistic model

by Jean-Cédric Chappelier -
Simply as you said: they are computed from the parameters.
If you have a function x\mapsto a\,x parametrized by a, the function, or even it's expression (a\,x), is not the same as its parameter (a).