<s>
The	O
Rocchio	B-Algorithm
algorithm	I-Algorithm
is	O
based	O
on	O
a	O
method	O
of	O
relevance	B-Application
feedback	I-Application
found	O
in	O
information	B-Library
retrieval	I-Library
systems	I-Library
which	O
stemmed	O
from	O
the	O
SMART	B-Application
Information	I-Application
Retrieval	I-Application
System	I-Application
developed	O
between	O
1960	O
and	O
1964	O
.	O
</s>
<s>
Like	O
many	O
other	O
retrieval	B-Library
systems	I-Library
,	O
the	O
Rocchio	B-Algorithm
algorithm	I-Algorithm
was	O
developed	O
using	O
the	O
vector	O
space	O
model	O
.	O
</s>
<s>
Therefore	O
,	O
the	O
user	O
's	O
search	O
query	B-Library
is	O
revised	O
to	O
include	O
an	O
arbitrary	O
percentage	O
of	O
relevant	O
and	O
irrelevant	O
documents	O
as	O
a	O
means	O
of	O
increasing	O
the	O
search	B-Application
engine	I-Application
's	O
recall	O
,	O
and	O
possibly	O
the	O
precision	O
as	O
well	O
.	O
</s>
<s>
The	O
number	O
of	O
relevant	O
and	O
irrelevant	O
documents	O
allowed	O
to	O
enter	O
a	O
query	B-Library
is	O
dictated	O
by	O
the	O
weights	O
of	O
the	O
a	O
,	O
b	O
,	O
c	O
variables	O
listed	O
below	O
in	O
the	O
Algorithm	O
section	O
.	O
</s>
<s>
The	O
formula	O
and	O
variable	O
definitions	O
for	O
Rocchio	O
relevance	B-Application
feedback	I-Application
are	O
as	O
follows	O
:	O
</s>
<s>
As	O
demonstrated	O
in	O
the	O
formula	O
,	O
the	O
associated	O
weights	O
(	O
a	O
,	O
b	O
,	O
c	O
)	O
are	O
responsible	O
for	O
shaping	O
the	O
modified	O
vector	O
in	O
a	O
direction	O
closer	O
,	O
or	O
farther	O
away	O
,	O
from	O
the	O
original	O
query	B-Library
,	O
related	O
documents	O
,	O
and	O
non-related	O
documents	O
.	O
</s>
<s>
If	O
the	O
user	O
decides	O
that	O
the	O
modified	O
query	B-Library
should	O
not	O
contain	O
terms	O
from	O
either	O
the	O
original	O
query	B-Library
,	O
related	O
documents	O
,	O
or	O
non-related	O
documents	O
,	O
then	O
the	O
corresponding	O
weight	O
(	O
a	O
,	O
b	O
,	O
c	O
)	O
value	O
for	O
the	O
category	O
should	O
be	O
set	O
to	O
0	O
.	O
</s>
<s>
In	O
the	O
later	O
part	O
of	O
the	O
algorithm	O
,	O
the	O
variables	O
,	O
and	O
are	O
presented	O
to	O
be	O
sets	O
of	O
vectors	B-Application
containing	O
the	O
coordinates	O
of	O
related	O
documents	O
and	O
non-related	O
documents	O
.	O
</s>
<s>
Though	O
and	O
are	O
not	O
vectors	B-Application
themselves	O
,	O
and	O
are	O
the	O
vectors	B-Application
used	O
to	O
iterate	O
through	O
the	O
two	O
sets	O
and	O
form	O
vector	O
summations	O
.	O
</s>
<s>
Thus	O
if	O
the	O
weight	O
is	O
increased	O
for	O
related	O
documents	O
,	O
then	O
the	O
modified	O
vectors	B-Application
coordinates	O
will	O
reflect	O
being	O
closer	O
to	O
the	O
centroid	O
of	O
related	O
documents	O
.	O
</s>
<s>
Therefore	O
,	O
traditional	O
values	O
for	O
the	O
algorithm	O
's	O
weights	O
(	O
a	O
,	O
b	O
,	O
c	O
)	O
in	O
Rocchio	B-Algorithm
classification	I-Algorithm
are	O
typically	O
around	O
a	O
=	O
1	O
,	O
b	O
=	O
0.8	O
,	O
and	O
c	O
=	O
0.1	O
.	O
</s>
<s>
Modern	O
information	B-Library
retrieval	I-Library
systems	I-Library
have	O
moved	O
towards	O
eliminating	O
the	O
non-related	O
documents	O
by	O
setting	O
c	O
=	O
0	O
and	O
thus	O
only	O
accounting	O
for	O
related	O
documents	O
.	O
</s>
<s>
Although	O
not	O
all	O
retrieval	B-Library
systems	I-Library
have	O
eliminated	O
the	O
need	O
for	O
non-related	O
documents	O
,	O
most	O
have	O
limited	O
the	O
effects	O
on	O
modified	O
query	B-Library
by	O
only	O
accounting	O
for	O
strongest	O
non-related	O
documents	O
in	O
the	O
set	O
.	O
</s>
<s>
The	O
Rocchio	B-Algorithm
algorithm	I-Algorithm
often	O
fails	O
to	O
classify	O
multimodal	O
classes	O
and	O
relationships	O
.	O
</s>
<s>
Therefore	O
,	O
the	O
two	O
queries	B-Library
of	O
"	O
Burma	O
"	O
and	O
"	O
Myanmar	O
"	O
will	O
appear	O
much	O
farther	O
apart	O
in	O
the	O
vector	O
space	O
model	O
,	O
though	O
they	O
both	O
contain	O
similar	O
origins	O
.	O
</s>
