20.07.2013 Views

Notes on computational linguistics.pdf - UCLA Department of ...

Notes on computational linguistics.pdf - UCLA Department of ...

Notes on computational linguistics.pdf - UCLA Department of ...

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

Stabler - Lx 185/209 2003<br />

(109) If the outcomes <strong>of</strong> the binary decisi<strong>on</strong>s are not equally likely, though, we want to say something else.<br />

The amount <strong>of</strong> informati<strong>on</strong> (or “self-informati<strong>on</strong>” or the “surprisal”) <strong>of</strong> an event A,<br />

i(A) = log<br />

1<br />

=−log P(A)<br />

P(A)<br />

So if we have 10 possible events with equal probabilities <strong>of</strong> occurrence, so P(A) = 0.1, then<br />

i(A) = log 1<br />

=−log 0.1 ≈ 3.32<br />

0.1<br />

(110) The simple cases still work out properly.<br />

In the easiest case where probability is distributed uniformly across 8 possibilities in ΩX, wewouldhave<br />

exactly 3 bits <strong>of</strong> informati<strong>on</strong> given by the occurrence <strong>of</strong> a particular event A:<br />

i(A) = log<br />

1<br />

=−log 0.125 = 3<br />

0.125<br />

The informati<strong>on</strong> given by the occurrence <strong>of</strong> ∪ΩX, whereP(∪ΩX) = 1, is zero:<br />

i(A) = log 1<br />

=−log 1 = 0<br />

1<br />

And obviously, if events A, B ∈ ΩX are independent, that is, P(AB) = P(A)P(B),then<br />

i(AB) = log<br />

1<br />

P(AB)<br />

1<br />

= log P(A)P(B)<br />

= log 1<br />

1<br />

P(A) + log P(B)<br />

= i(A) + i(B)<br />

(111) However, in the case where ΩX ={A, B} where P(A) = 0.1 andP(B) = 0.9, we will still have<br />

i(A) = log 1<br />

=−log 0.1 ≈ 3.32<br />

0.1<br />

That is, this event c<strong>on</strong>veys more than 3 bits <strong>of</strong> informati<strong>on</strong> even though there is <strong>on</strong>ly <strong>on</strong>e other opti<strong>on</strong>.<br />

The informati<strong>on</strong> c<strong>on</strong>veyed by the other event<br />

i(B) = log 1<br />

≈ .15<br />

0.9<br />

155

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!