W3 Quiz Q5 answer wrong?

Course Q&A Natural Language Processing NLP with Probabilistic Models
1

I have attempted Week 3 Quiz and can’t understand the correct answer for Q5 (unanswered screenshot below). Is the correct answer referring to the wrong m? Thanks.

image

2

Hey, while I can’t give you the right answer, you just have to use the formula for perplexity provided in the lecture and reading material.
log PP(W) = (-1/m)*sum(log_2(P(w_i|w_i-1))

3

Thank you for the pointer. My main confusion is that m = 5 in this case right? (Which is not what the correct answer says…)

4

I don’t think it’s explained, but the slide with the cross-eyed penguin says we count and do not count .

5

Vincent’s notation for the start and end sequences in his answer below didn’t get through (probably interpreted as HTML tags), so for others wondering this same question as I was…

Perplexity includes the end-of-sentence tokens but not the start-of-sentence tokens.

I think this is because the problem isn’t symmetric: we are predicting likely next words… so every sentence has to start with a start token – there’s no predictive value there. Versus some words are statistically more likely than others to end a sentence (thus have predictive value of the next word (end token)).