<s>
Product	B-Application
finders	I-Application
are	O
information	O
systems	O
that	O
help	O
consumers	O
to	O
identify	O
products	O
within	O
a	O
large	O
palette	O
of	O
similar	O
alternative	O
products	O
.	O
</s>
<s>
Product	B-Application
finders	I-Application
differ	O
in	O
complexity	O
,	O
the	O
more	O
complex	O
among	O
them	O
being	O
a	O
special	O
case	O
of	O
decision	B-Application
support	I-Application
systems	I-Application
.	O
</s>
<s>
Conventional	O
decision	B-Application
support	I-Application
systems	I-Application
,	O
however	O
,	O
aim	O
at	O
specialized	O
user	O
groups	O
,	O
e.g.	O
</s>
<s>
marketing	O
managers	O
,	O
whereas	O
product	B-Application
finders	I-Application
focus	O
on	O
consumers	O
.	O
</s>
<s>
Usually	O
,	O
product	B-Application
finders	I-Application
are	O
part	O
of	O
an	O
e-shop	O
or	O
an	O
online	O
presentation	O
of	O
a	O
product-line	O
.	O
</s>
<s>
Being	O
part	O
of	O
an	O
e-shop	O
,	O
a	O
product	B-Application
finder	I-Application
ideally	O
leads	O
to	O
an	O
online	O
buy	O
,	O
while	O
conventional	O
distribution	O
channels	O
are	O
involved	O
in	O
product	B-Application
finders	I-Application
that	O
are	O
part	O
of	O
an	O
online	O
presentation	O
(	O
e.g.	O
</s>
<s>
Product	B-Application
finders	I-Application
are	O
best	O
suited	O
for	O
product	O
groups	O
whose	O
individual	O
products	O
are	O
comparable	O
by	O
specific	O
criteria	O
.	O
</s>
<s>
This	O
is	O
true	O
,	O
in	O
most	O
cases	O
,	O
with	O
technical	O
products	O
such	O
as	O
notebooks	B-Device
:	O
their	O
features	O
(	O
e.g.	O
</s>
<s>
clock	O
rate	O
,	O
size	O
of	O
harddisk	B-Device
,	O
price	O
,	O
screen	O
size	O
)	O
may	O
influence	O
the	O
consumer	O
's	O
decision	O
.	O
</s>
<s>
Beside	O
technical	O
products	O
such	O
as	O
notebooks	B-Device
,	O
cars	O
,	O
dish	O
washers	O
,	O
cell	O
phones	O
or	O
GPS	O
devices	O
,	O
non-technical	O
products	O
such	O
as	O
wine	O
,	O
socks	O
,	O
toothbrushes	O
or	O
nails	O
may	O
be	O
supported	O
by	O
product	B-Application
finders	I-Application
as	O
well	O
,	O
as	O
comparison	O
by	O
features	O
takes	O
place	O
.	O
</s>
<s>
On	O
the	O
other	O
hand	O
,	O
the	O
application	O
of	O
product	B-Application
finders	I-Application
is	O
limited	O
when	O
it	O
comes	O
to	O
individualized	O
products	O
such	O
as	O
books	O
,	O
jewelry	O
or	O
compact	O
discs	O
as	O
consumers	O
do	O
not	O
select	O
such	O
products	O
along	O
specific	O
,	O
comparable	O
features	O
.	O
</s>
<s>
Furthermore	O
,	O
product	B-Application
finders	I-Application
are	O
used	O
not	O
only	O
for	O
products	O
sensu	O
stricto	O
,	O
but	O
for	O
services	O
as	O
well	O
,	O
e.g.	O
</s>
<s>
Product	B-Application
finders	I-Application
are	O
used	O
both	O
by	O
manufacturers	O
,	O
dealers	O
(	O
comprising	O
several	O
manufacturers	O
)	O
,	O
and	O
web	O
portals	O
(	O
comprising	O
several	O
dealers	O
)	O
.	O
</s>
<s>
There	O
is	O
a	O
move	O
to	O
integrate	O
Product	B-Application
finders	I-Application
with	O
social	O
networking	O
and	O
group	O
buying	O
allowing	O
users	O
to	O
add	O
and	O
rate	O
products	O
,	O
locations	O
and	O
purchase	O
recommended	O
products	O
with	O
others	O
.	O
</s>
<s>
Dialogue	B-General_Concept
systems	I-General_Concept
or	O
Interactive	O
product	B-Application
finders	I-Application
(	O
Product	O
Wizards	O
)	O
–	O
Interactive	O
Product	B-Application
finders	I-Application
are	O
dialogue-based	O
recommendation	O
solutions	O
that	O
provide	O
shoppers	O
with	O
personalized	O
,	O
need-oriented	O
support	O
as	O
they	O
want	O
to	O
choose	O
the	O
right	O
product	O
.	O
</s>
<s>
Comparison	O
table	O
–	O
A	O
comparison	O
table	O
is	O
a	O
basic	O
version	O
of	O
a	O
product	B-Application
finder	I-Application
that	O
allows	O
consumers	O
to	O
easily	O
compare	O
products	O
,	O
features	O
and	O
prices	O
.	O
</s>
<s>
String	B-Algorithm
search	I-Algorithm
–	O
A	O
string	B-Algorithm
search	I-Algorithm
algorithm	I-Algorithm
locates	O
where	O
several	O
smaller	O
strings	O
are	O
within	O
a	O
larger	O
text	O
.	O
</s>
<s>
For	O
example	O
,	O
if	O
a	O
user	O
typed	O
"	O
smart	O
phone	O
"	O
into	O
a	O
Google	B-Application
search	I-Application
,	O
Google	B-Application
would	O
be	O
searching	O
to	O
find	O
where	O
that	O
keyword	O
is	O
located	O
within	O
different	O
scripts	O
and	O
codes	O
to	O
refer	O
the	O
user	O
to	O
the	O
most	O
relevant	O
information	O
possible	O
.	O
</s>
<s>
Filtering	O
systems	O
–	O
An	O
information	B-Application
filtering	I-Application
system	I-Application
is	O
a	O
system	O
that	O
removes	O
redundant	O
information	O
from	O
an	O
information	O
stream	O
before	O
presenting	O
it	O
to	O
a	O
human	O
user	O
.	O
</s>
<s>
A	O
notebook	B-Device
filter	O
,	O
for	O
instance	O
,	O
allows	O
users	O
to	O
select	O
features	O
to	O
narrow	O
down	O
the	O
list	O
of	O
displayed	O
products	O
.	O
</s>
<s>
Scoring	O
systems	O
–	O
Scoring	O
systems	O
are	O
often	O
found	O
on	O
recommender	B-Application
systems	I-Application
and	O
allow	O
users	O
to	O
rate	O
products	O
for	O
other	O
users	O
to	O
see	O
.	O
</s>
<s>
The	O
Mac	O
Observer	O
,	O
a	O
popular	O
recommender	B-Application
and	O
news	O
site	O
that	O
reviews	O
Apple	O
products	O
,	O
has	O
recently	O
announced	O
they	O
will	O
be	O
changing	O
their	O
scoring	O
system	O
.	O
</s>
<s>
Tagging	O
clouds	O
–	O
A	O
tag	B-Application
cloud	I-Application
is	O
a	O
visual	O
representation	O
of	O
text	O
data	O
,	O
used	O
to	O
simplified	O
and	O
decode	O
keywords	O
and	O
tags	B-General_Concept
on	O
websites	O
.	O
</s>
<s>
The	O
tags	B-General_Concept
are	O
usually	O
single	O
words	O
and	O
the	O
importance	O
of	O
each	O
tag	B-General_Concept
is	O
represented	O
by	O
the	O
color	O
and	O
size	O
of	O
the	O
word	O
.	O
</s>
<s>
In	O
product	B-Application
finders	I-Application
,	O
tag	B-Application
clouds	I-Application
will	O
have	O
their	O
tags	B-General_Concept
hyperlinked	O
so	O
that	O
a	O
user	O
can	O
easily	O
navigate	O
the	O
website	O
.	O
</s>
<s>
To	O
find	O
the	O
product	O
the	O
user	O
is	O
looking	O
for	O
,	O
they	O
would	O
find	O
the	O
tag	B-General_Concept
within	O
the	O
cloud	O
,	O
click	O
on	O
the	O
tag	B-General_Concept
and	O
be	O
directed	O
to	O
a	O
landing	O
page	O
where	O
their	O
desired	O
product	O
is	O
featured	O
.	O
</s>
<s>
Neural	B-General_Concept
Networks	I-General_Concept
–	O
A	O
neural	B-General_Concept
network	I-General_Concept
is	O
a	O
family	O
of	O
learning	O
models	O
inspired	O
by	O
biological	B-General_Concept
neural	I-General_Concept
networks	I-General_Concept
(	O
the	O
nervous	O
systems	O
of	O
animals	O
,	O
in	O
particular	O
the	O
brain	O
)	O
and	O
are	O
used	O
to	O
estimate	O
user	O
preferences	O
.	O
</s>
<s>
Neural	B-General_Concept
networks	I-General_Concept
have	O
classification	B-General_Concept
abilities	O
,	O
including	O
pattern	O
recognition	O
.	O
</s>
<s>
Netflix	O
,	O
for	O
example	O
,	O
uses	O
a	O
neural	B-General_Concept
network	I-General_Concept
to	O
see	O
what	O
genre	O
of	O
movies	O
you	O
prefer	O
to	O
watch	O
.	O
</s>
<s>
Neural	B-General_Concept
networks	I-General_Concept
also	O
do	O
data	O
processing	O
,	O
including	O
data	O
filtering	O
,	O
similar	O
to	O
the	O
purpose	O
of	O
a	O
filtering	O
system	O
.	O
</s>
<s>
Relational	B-Application
Database	I-Application
–	O
A	O
relational	B-Application
database	I-Application
is	O
a	O
digital	O
database	O
which	O
organizes	O
data	O
into	O
tables	O
(	O
or	O
"	O
relations	O
"	O
)	O
of	O
rows	O
and	O
columns	O
,	O
with	O
a	O
unique	O
key	O
for	O
each	O
row	O
.	O
</s>
<s>
Unlike	O
hierarchical	O
tables	O
such	O
as	O
menu	O
trees	O
,	O
relational	B-Application
database	I-Application
tables	O
can	O
have	O
rows	O
that	O
are	O
linked	O
to	O
rows	O
in	O
other	O
tables	O
by	O
a	O
keyword	O
that	O
they	O
may	O
share	O
.	O
</s>
<s>
Databases	O
like	O
these	O
make	O
it	O
simple	O
for	O
product	B-Application
finders	I-Application
to	O
discover	O
the	O
relationships	O
between	O
keywords	O
that	O
consumer	O
uses	O
.	O
</s>
<s>
Product	B-Application
finder	I-Application
has	O
an	O
important	O
role	O
in	O
e-commerce	O
,	O
items	O
has	O
to	O
be	O
categorized	O
to	O
better	O
serve	O
consumer	O
in	O
searching	O
the	O
desired	O
product	O
,	O
recommender	B-Application
system	I-Application
for	O
recommending	O
items	O
based	O
on	O
their	O
purchases	O
etc	O
.	O
</s>
<s>
Large	O
online	O
consumer	O
to	O
consumer	O
marketplaces	O
such	O
as	O
eBay	B-Application
,	O
Amazon	B-Application
,	O
and	O
Alibaba	O
feature	O
millions	O
of	O
items	O
with	O
more	O
entered	O
into	O
the	O
marketplace	O
every	O
day	O
.	O
</s>
<s>
Item	O
categorization	O
helps	O
in	O
classifying	O
products	O
and	O
giving	O
them	O
tags	B-General_Concept
and	O
labels	O
,	O
which	O
helps	O
consumer	O
find	O
them	O
.	O
</s>
<s>
Traditionally	O
bag-of-words	B-General_Concept
model	I-General_Concept
approach	O
is	O
used	O
to	O
solve	O
the	O
problem	O
with	O
using	O
no	O
hierarchy	O
at	O
all	O
or	O
using	O
human-defined	O
hierarchy	O
.	O
</s>
<s>
A	O
new	O
method	O
,	O
using	O
hierarchical	O
approach	O
which	O
decomposes	O
the	O
classification	B-General_Concept
problem	O
into	O
a	O
coarse	O
level	O
task	O
and	O
a	O
fine	O
level	O
task	O
,	O
with	O
the	O
hierarchy	O
made	O
using	O
latent	O
class	O
model	O
discovery	O
.	O
</s>
<s>
A	O
simple	O
classifier	B-General_Concept
is	O
applied	O
to	O
perform	O
the	O
coarse	O
level	O
classification	B-General_Concept
(	O
because	O
the	O
data	O
is	O
so	O
large	O
we	O
cannot	O
use	O
more	O
sophisticated	O
approach	O
due	O
to	O
time	O
issue	O
)	O
while	O
a	O
more	O
sophisticated	O
model	O
is	O
used	O
to	O
separate	O
classes	O
at	O
the	O
fine	O
level	O
.	O
</s>
<s>
Then	O
we	O
form	O
a	O
confusion	B-General_Concept
matrix	I-General_Concept
between	O
groups	O
to	O
approximate	O
the	O
similarity	O
of	O
classes	O
,	O
the	O
similar	O
classes	O
are	O
kept	O
in	O
a	O
group	O
and	O
so	O
at	O
every	O
stage	O
we	O
get	O
groups	O
with	O
no	O
similarity	O
and	O
hence	O
we	O
get	O
a	O
hierarchy	O
tree	O
.	O
</s>
<s>
At	O
Coarse	O
level	O
we	O
classify	O
the	O
testing	O
instance	O
,	O
for	O
one	O
of	O
the	O
groups	O
at	O
the	O
first	O
level	O
of	O
hierarchy	O
,	O
As	O
the	O
data	O
set	O
is	O
large	O
we	O
cannot	O
use	O
sophisticated	O
algorithm	O
,	O
and	O
thus	O
at	O
this	O
stage	O
either	O
KNN	B-General_Concept
or	O
Naive	B-General_Concept
Bayes	I-General_Concept
is	O
used	O
.	O
</s>
<s>
At	O
fine	O
level	O
we	O
classify	O
the	O
items	O
within	O
a	O
group	O
into	O
some	O
subset	O
group	O
,	O
as	O
there	O
can	O
be	O
similarity	O
in	O
the	O
group	O
we	O
use	O
a	O
sophisticated	O
mechanism	O
,	O
generally	O
SVM	B-Algorithm
at	O
every	O
node	O
.	O
</s>
<s>
KNN	B-General_Concept
(	O
k	O
nearest	O
neighbours	O
)	O
algorithm	O
finds	O
the	O
k	O
neighbours	O
which	O
are	O
really	O
similar	O
to	O
the	O
testing	O
instance	O
,	O
it	O
uses	O
Euclidean	O
or	O
cosine	O
similarity	O
function	O
to	O
find	O
the	O
distance	O
between	O
each	O
class	O
and	O
then	O
gives	O
the	O
top	O
k	O
class	O
.	O
</s>
<s>
In	O
this	O
example	O
the	O
coarse	O
grained	O
classifier	B-General_Concept
would	O
tell	O
us	O
that	O
the	O
testing	O
instance	O
belongs	O
to	O
electronic	O
group	O
,	O
then	O
we	O
use	O
fine	O
grained	O
at	O
every	O
stage	O
and	O
we	O
got	O
this	O
tree	O
.	O
</s>
<s>
Recommendation	B-Application
systems	I-Application
are	O
used	O
to	O
recommend	O
consumer	O
items/product	O
based	O
on	O
their	O
purchasing	O
or	O
search	O
history	O
.	O
</s>
