<s>
Item-item	B-Application
collaborative	I-Application
filtering	I-Application
,	O
or	O
item-based	O
,	O
or	O
item-to-item	O
,	O
is	O
a	O
form	O
of	O
collaborative	B-Algorithm
filtering	I-Algorithm
for	O
recommender	B-Application
systems	I-Application
based	O
on	O
the	O
similarity	O
between	O
items	O
calculated	O
using	O
people	O
's	O
ratings	O
of	O
those	O
items	O
.	O
</s>
<s>
Item-item	B-Application
collaborative	I-Application
filtering	I-Application
was	O
invented	O
and	O
used	O
by	O
Amazon.com	B-Application
in	O
1998	O
.	O
</s>
<s>
Earlier	O
collaborative	B-Algorithm
filtering	I-Algorithm
systems	O
based	O
on	O
rating	O
similarity	O
between	O
users	O
(	O
known	O
as	O
user-user	O
collaborative	B-Algorithm
filtering	I-Algorithm
)	O
had	O
several	O
problems	O
:	O
</s>
<s>
Second	O
,	O
the	O
system	O
executes	O
a	O
recommendation	B-Application
stage	O
.	O
</s>
<s>
Usually	O
this	O
calculation	O
is	O
a	O
weighted	B-Algorithm
sum	I-Algorithm
or	O
linear	B-General_Concept
regression	I-General_Concept
.	O
</s>
<s>
This	O
form	O
of	O
recommendation	B-Application
is	O
analogous	O
to	O
"	O
people	O
who	O
rate	O
item	O
X	O
highly	O
,	O
like	O
you	O
,	O
also	O
tend	O
to	O
rate	O
item	O
Y	O
highly	O
,	O
and	O
you	O
have	O
n't	O
rated	O
item	O
Y	O
yet	O
,	O
so	O
you	O
should	O
try	O
it	O
"	O
.	O
</s>
<s>
Item-item	B-Application
collaborative	I-Application
filtering	I-Application
had	O
less	O
error	O
than	O
user-user	O
collaborative	B-Algorithm
filtering	I-Algorithm
.	O
</s>
