lecture18-learning-ranking-handouts-6-per

G the characters in the query 2 53011

Info iconThis preview shows page 1. Sign up to view the full content.

View Full Document Right Arrow Icon
This is the end of the preview. Sign up to access the rest of the document.

Unformatted text preview: 
 URL
contains
“~”?
 Page
edit
recency?
 Page
length?
   Well
compute
a
score
in
[0,1]
for
each
doc
d and
for
 each
query
q
using
a
linear
combina)on
of
sT
and
sB
   Major
web
search
engines
publicly
state
that
they
use
 “hundreds”
of
such
features
–
and
they
keep
changing
 Introduc)on to Informa)on Retrieval Sec.
6.1.2
 Thus
our
scores
are
all
0,
g, 1‐g or
1.
   g is
a
parameter
to
be
learned
from
examples Introduc)on to Informa)on Retrieval Sec.
6.1.2
 We
are
given
examples
 Least
square
errors
   Created
by
human
judges
   For
each
human‐judged
example,
we
compute
its
 squared
error:
   Then
we
can
compute
a
total
error
of
   We
quan)ze
the
human
relevance
judgments
to
be
1
 or
0
respec)vely,
for
Relevant
and
Non‐relevant
   We
will
pick
g to
minimize
this
total
error.
   The
scores
we
compute
will
be
0,
g, 1‐g or
1
...
View Full Document

Ask a homework question - tutors are online