https://miro.com/app/board/o9l IIS4AVg=/
Relevance feedback + Text classification (Chapter 9+13)
Definition 1 (Rocchlo relevance feedback)
Rocchio relevance feedback has the form
initial
query /'
= ago + ß
where qa is the original query vector, Dr is the set of relevant documents, Dnr is the set of non-relevant documents and the values a, 0, 7 depend on the system setting.
Exercise 9/1
What is the main purpose of Rocchio relevance feedback?
Revised query
x known non-relevant documents
Exercise 9/2
A user's primary query i^ehcap CDs cheap DVDs extremely chcapCD&)Thc user has ;i In ok on I wn don^nonls; doc I a dor'2. rn,r'k in;1; doe: I ("■/"/,■ cheap sojhuo.ir c.iicap CDs ay relevant iuji; doc'2 'leap /.h'-ill'i D\ D.i at; 11011-re levant. Assume that we use a simple tf scheme, without vp.c~.nr length tio"iri;ili/.;i.tiori. What wniild! hp. I.hp restructured query vector- after e<ui:uder_n;: tin:.' Ren ol:i;> it.'l:.;vaii:;c Ice; 11;ark with values a = 1, 0 = 0.75, ar.d 7 = 0.25?
Text classification and Naive Bayes (Chapter 13)
Definition 2 (Naive Bayes Classifier)
Naive Bayes (NB) Clussiji.tr assumes l.h.al. Iht effect of the value of a predictor x on a given class c is class co-ndilio-nal imltpendt.nl.. Bayes t.h.eorem. providts a way of calculating Iht. ■posterior probability P[<: .r) from class prior proim.irhi.tij P'c). predictor prior probability P{x) and probability of the predictor given the class P(x\c)
Pics.)
PiAc)P{c) P{x)
and for a vector of predictors X =
P(c\X) =
P(x1\c)...P(xn\c)P(c)
P(Xl)...P(xn)
The class with the highest posterior probability is the outcome of prediction.
Wc rewrite the
to the tabic for a
	rčíc van t	non-relevant	
terms	docl	doc2	query
CDs	2	0	2
cheap	2	1	3
software	1	0	0
thrills	0	1	0
DVDs	0	1	1
■; ;xl. re in ch-	0	0	1