<s>
Search	B-Application
engine	I-Application
indexing	I-Application
is	O
the	O
collecting	O
,	O
parsing	B-Language
,	O
and	O
storing	O
of	O
data	O
to	O
facilitate	O
fast	O
and	O
accurate	O
information	B-Library
retrieval	I-Library
.	O
</s>
<s>
Index	O
design	O
incorporates	O
interdisciplinary	O
concepts	O
from	O
linguistics	O
,	O
cognitive	O
psychology	O
,	O
mathematics	O
,	O
informatics	O
,	O
and	O
computer	B-General_Concept
science	I-General_Concept
.	O
</s>
<s>
An	O
alternate	O
name	O
for	O
the	O
process	O
,	O
in	O
the	O
context	O
of	O
search	B-Application
engines	I-Application
designed	O
to	O
find	O
web	O
pages	O
on	O
the	O
Internet	O
,	O
is	O
web	O
indexing	O
.	O
</s>
<s>
Popular	O
engines	O
focus	O
on	O
the	O
full-text	B-Application
indexing	O
of	O
online	O
,	O
natural	B-Language
language	I-Language
documents	O
.	O
</s>
<s>
Media	B-Application
types	I-Application
such	O
as	O
pictures	O
,	O
video	O
,	O
audio	O
,	O
and	O
graphics	O
are	O
also	O
searchable	B-Application
.	O
</s>
<s>
Meta	B-Application
search	I-Application
engines	I-Application
reuse	O
the	O
indices	O
of	O
other	O
services	O
and	O
do	O
not	O
store	O
a	O
local	O
index	O
whereas	O
cache-based	O
search	B-Application
engines	I-Application
permanently	O
store	O
the	O
index	O
along	O
with	O
the	O
corpus	O
.	O
</s>
<s>
Unlike	O
full-text	B-Application
indices	O
,	O
partial-text	O
services	O
restrict	O
the	O
depth	O
indexed	O
to	O
reduce	O
index	O
size	O
.	O
</s>
<s>
Larger	O
services	O
typically	O
perform	O
indexing	O
at	O
a	O
predetermined	O
time	O
interval	O
due	O
to	O
the	O
required	O
time	O
and	O
processing	O
costs	O
,	O
while	O
agent-based	O
search	B-Application
engines	I-Application
index	O
in	O
real	B-General_Concept
time	I-General_Concept
.	O
</s>
<s>
The	O
purpose	O
of	O
storing	O
an	O
index	O
is	O
to	O
optimize	O
speed	O
and	O
performance	O
in	O
finding	O
relevant	O
documents	O
for	O
a	O
search	O
query	B-Library
.	O
</s>
<s>
Without	O
an	O
index	O
,	O
the	O
search	B-Application
engine	I-Application
would	O
scan	B-Application
every	O
document	O
in	O
the	O
corpus	O
,	O
which	O
would	O
require	O
considerable	O
time	O
and	O
computing	O
power	O
.	O
</s>
<s>
For	O
example	O
,	O
while	O
an	O
index	O
of	O
10,000	O
documents	O
can	O
be	O
queried	O
within	O
milliseconds	O
,	O
a	O
sequential	O
scan	B-Application
of	O
every	O
word	O
in	O
10,000	O
large	O
documents	O
could	O
take	O
hours	O
.	O
</s>
<s>
The	O
additional	O
computer	B-General_Concept
storage	I-General_Concept
required	O
to	O
store	O
the	O
index	O
,	O
as	O
well	O
as	O
the	O
considerable	O
increase	O
in	O
the	O
time	O
required	O
for	O
an	O
update	O
to	O
take	O
place	O
,	O
are	O
traded	O
off	O
for	O
the	O
time	O
saved	O
during	O
information	B-Library
retrieval	I-Library
.	O
</s>
<s>
Major	O
factors	O
in	O
designing	O
a	O
search	B-Application
engine	I-Application
's	O
architecture	O
include	O
:	O
</s>
<s>
Merge	B-Language
factors	O
How	O
data	O
enters	O
the	O
index	O
,	O
or	O
how	O
words	O
or	O
subject	O
features	O
are	O
added	O
to	O
the	O
index	O
during	O
text	O
corpus	O
traversal	O
,	O
and	O
whether	O
multiple	O
indexers	O
can	O
work	O
asynchronously	O
.	O
</s>
<s>
Traversal	O
typically	O
correlates	O
to	O
the	O
data	B-Application
collection	I-Application
policy	O
.	O
</s>
<s>
Search	B-Application
engine	I-Application
index	O
merging	O
is	O
similar	O
in	O
concept	O
to	O
the	O
SQL	B-Language
Merge	I-Language
command	O
and	O
other	O
merge	B-Language
algorithms	O
.	O
</s>
<s>
Storage	O
techniques	O
How	O
to	O
store	O
the	O
index	O
data	O
,	O
that	O
is	O
,	O
whether	O
information	O
should	O
be	O
data	O
compressed	B-Operating_System
or	O
filtered	O
.	O
</s>
<s>
Index	O
size	O
How	O
much	O
computer	B-General_Concept
storage	I-General_Concept
is	O
required	O
to	O
support	O
the	O
index	O
.	O
</s>
<s>
Lookup	O
speed	O
How	O
quickly	O
a	O
word	O
can	O
be	O
found	O
in	O
the	O
inverted	B-Algorithm
index	I-Algorithm
.	O
</s>
<s>
The	O
speed	O
of	O
finding	O
an	O
entry	O
in	O
a	O
data	O
structure	O
,	O
compared	O
with	O
how	O
quickly	O
it	O
can	O
be	O
updated	O
or	O
removed	O
,	O
is	O
a	O
central	O
focus	O
of	O
computer	B-General_Concept
science	I-General_Concept
.	O
</s>
<s>
Issues	O
include	O
dealing	O
with	O
index	O
corruption	O
,	O
determining	O
whether	O
bad	O
data	O
can	O
be	O
treated	O
in	O
isolation	O
,	O
dealing	O
with	O
bad	O
hardware	B-Architecture
,	O
partitioning	B-General_Concept
,	O
and	O
schemes	O
such	O
as	O
hash-based	B-Error_Name
or	O
composite	O
partitioning	B-General_Concept
,	O
as	O
well	O
as	O
replication	B-Architecture
.	O
</s>
<s>
Search	B-Application
engine	I-Application
architectures	O
vary	O
in	O
the	O
way	O
indexing	O
is	O
performed	O
and	O
in	O
methods	O
of	O
index	O
storage	O
to	O
meet	O
the	O
various	O
design	O
factors	O
.	O
</s>
<s>
Suffix	B-Architecture
tree	I-Architecture
Figuratively	O
structured	O
like	O
a	O
tree	O
,	O
supports	O
linear	O
time	O
lookup	O
.	O
</s>
<s>
The	O
suffix	B-Architecture
tree	I-Architecture
is	O
a	O
type	O
of	O
trie	B-General_Concept
.	O
</s>
<s>
Tries	B-General_Concept
support	O
extendible	B-Algorithm
hashing	I-Algorithm
,	O
which	O
is	O
important	O
for	O
search	B-Application
engine	I-Application
indexing	I-Application
.	O
</s>
<s>
An	O
alternate	O
representation	O
is	O
a	O
suffix	B-Algorithm
array	I-Algorithm
,	O
which	O
is	O
considered	O
to	O
require	O
less	O
virtual	O
memory	O
and	O
supports	O
data	B-General_Concept
compression	I-General_Concept
such	O
as	O
the	O
BWT	B-Algorithm
algorithm	O
.	O
</s>
<s>
Inverted	B-Algorithm
index	I-Algorithm
Stores	O
a	O
list	O
of	O
occurrences	O
of	O
each	O
atomic	O
search	O
criterion	O
,	O
typically	O
in	O
the	O
form	O
of	O
a	O
hash	B-Algorithm
table	I-Algorithm
or	O
binary	O
tree	O
.	O
</s>
<s>
n-gram	B-Language
index	O
Stores	O
sequences	O
of	O
length	O
of	O
data	O
to	O
support	O
other	O
types	O
of	O
retrieval	O
or	O
text	B-Algorithm
mining	I-Algorithm
.	O
</s>
<s>
Document-term	B-General_Concept
matrix	I-General_Concept
Used	O
in	O
latent	O
semantic	O
analysis	O
,	O
stores	O
the	O
occurrences	O
of	O
words	O
in	O
documents	O
in	O
a	O
two-dimensional	O
sparse	B-Algorithm
matrix	I-Algorithm
.	O
</s>
<s>
A	O
major	O
challenge	O
in	O
the	O
design	O
of	O
search	B-Application
engines	I-Application
is	O
the	O
management	O
of	O
serial	O
computing	O
processes	O
.	O
</s>
<s>
There	O
are	O
many	O
opportunities	O
for	O
race	B-Operating_System
conditions	I-Operating_System
and	O
coherent	O
faults	O
.	O
</s>
<s>
For	O
example	O
,	O
a	O
new	O
document	O
is	O
added	O
to	O
the	O
corpus	O
and	O
the	O
index	O
must	O
be	O
updated	O
,	O
but	O
the	O
index	O
simultaneously	O
needs	O
to	O
continue	O
responding	O
to	O
search	O
queries	B-Library
.	O
</s>
<s>
Consider	O
that	O
authors	O
are	O
producers	O
of	O
information	O
,	O
and	O
a	O
web	B-Application
crawler	I-Application
is	O
the	O
consumer	O
of	O
this	O
information	O
,	O
grabbing	O
the	O
text	O
and	O
storing	O
it	O
in	O
a	O
cache	O
(	O
or	O
corpus	O
)	O
.	O
</s>
<s>
The	O
forward	O
index	O
is	O
the	O
consumer	O
of	O
the	O
information	O
produced	O
by	O
the	O
corpus	O
,	O
and	O
the	O
inverted	B-Algorithm
index	I-Algorithm
is	O
the	O
consumer	O
of	O
information	O
produced	O
by	O
the	O
forward	O
index	O
.	O
</s>
<s>
The	O
indexer	O
is	O
the	O
producer	O
of	O
searchable	B-Application
information	O
and	O
users	O
are	O
the	O
consumers	O
that	O
need	O
to	O
search	O
.	O
</s>
<s>
The	O
challenge	O
is	O
magnified	O
when	O
working	O
with	O
distributed	O
storage	O
and	O
distributed	B-Architecture
processing	I-Architecture
.	O
</s>
<s>
In	O
an	O
effort	O
to	O
scale	O
with	O
larger	O
amounts	O
of	O
indexed	O
information	O
,	O
the	O
search	B-Application
engine	I-Application
's	O
architecture	O
may	O
involve	O
distributed	B-Architecture
computing	I-Architecture
,	O
where	O
the	O
search	B-Application
engine	I-Application
consists	O
of	O
several	O
machines	O
operating	O
in	O
unison	O
.	O
</s>
<s>
Many	O
search	B-Application
engines	I-Application
incorporate	O
an	O
inverted	B-Algorithm
index	I-Algorithm
when	O
evaluating	O
a	O
search	O
query	B-Library
to	O
quickly	O
locate	O
documents	O
containing	O
the	O
words	O
in	O
a	O
query	B-Library
and	O
then	O
rank	O
these	O
documents	O
by	O
relevance	O
.	O
</s>
<s>
Because	O
the	O
inverted	B-Algorithm
index	I-Algorithm
stores	O
a	O
list	O
of	O
the	O
documents	O
containing	O
each	O
word	O
,	O
the	O
search	B-Application
engine	I-Application
can	O
use	O
direct	O
access	B-General_Concept
to	O
find	O
the	O
documents	O
associated	O
with	O
each	O
word	O
in	O
the	O
query	B-Library
in	O
order	O
to	O
retrieve	O
the	O
matching	O
documents	O
quickly	O
.	O
</s>
<s>
The	O
following	O
is	O
a	O
simplified	O
illustration	O
of	O
an	O
inverted	B-Algorithm
index	I-Algorithm
:	O
</s>
<s>
Such	O
an	O
index	O
determines	O
which	O
documents	O
match	O
a	O
query	B-Library
but	O
does	O
not	O
rank	O
matched	O
documents	O
.	O
</s>
<s>
Position	O
information	O
enables	O
the	O
search	O
algorithm	O
to	O
identify	O
word	O
proximity	O
to	O
support	O
searching	O
for	O
phrases	O
;	O
frequency	O
can	O
be	O
used	O
to	O
help	O
in	O
ranking	O
the	O
relevance	O
of	O
documents	O
to	O
the	O
query	B-Library
.	O
</s>
<s>
Such	O
topics	O
are	O
the	O
central	O
research	O
focus	O
of	O
information	B-Library
retrieval	I-Library
.	O
</s>
<s>
The	O
inverted	B-Algorithm
index	I-Algorithm
is	O
a	O
sparse	B-Algorithm
matrix	I-Algorithm
,	O
since	O
not	O
all	O
words	O
are	O
present	O
in	O
each	O
document	O
.	O
</s>
<s>
To	O
reduce	O
computer	B-General_Concept
storage	I-General_Concept
memory	O
requirements	O
,	O
it	O
is	O
stored	O
differently	O
from	O
a	O
two	O
dimensional	O
array	B-Data_Structure
.	O
</s>
<s>
The	O
index	O
is	O
similar	O
to	O
the	O
term	B-General_Concept
document	I-General_Concept
matrices	I-General_Concept
employed	O
by	O
latent	O
semantic	O
analysis	O
.	O
</s>
<s>
The	O
inverted	B-Algorithm
index	I-Algorithm
can	O
be	O
considered	O
a	O
form	O
of	O
a	O
hash	B-Algorithm
table	I-Algorithm
.	O
</s>
<s>
In	O
larger	O
indices	O
the	O
architecture	O
is	O
typically	O
a	O
distributed	B-Operating_System
hash	I-Operating_System
table	I-Operating_System
.	O
</s>
<s>
The	O
inverted	B-Algorithm
index	I-Algorithm
is	O
filled	O
via	O
a	O
merge	B-Language
or	O
rebuild	O
.	O
</s>
<s>
A	O
rebuild	O
is	O
similar	O
to	O
a	O
merge	B-Language
but	O
first	O
deletes	O
the	O
contents	O
of	O
the	O
inverted	B-Algorithm
index	I-Algorithm
.	O
</s>
<s>
The	O
architecture	O
may	O
be	O
designed	O
to	O
support	O
incremental	O
indexing	O
,	O
where	O
a	O
merge	B-Language
identifies	O
the	O
document	O
or	O
documents	O
to	O
be	O
added	O
or	O
updated	O
and	O
then	O
parses	B-Language
each	O
document	O
into	O
words	O
.	O
</s>
<s>
For	O
technical	O
accuracy	O
,	O
a	O
merge	B-Language
conflates	O
newly	O
indexed	O
documents	O
,	O
typically	O
residing	O
in	O
virtual	O
memory	O
,	O
with	O
the	O
index	O
cache	O
residing	O
on	O
one	O
or	O
more	O
computer	O
hard	O
drives	O
.	O
</s>
<s>
After	O
parsing	B-Language
,	O
the	O
indexer	O
adds	O
the	O
referenced	O
document	O
to	O
the	O
document	O
list	O
for	O
the	O
appropriate	O
words	O
.	O
</s>
<s>
In	O
a	O
larger	O
search	B-Application
engine	I-Application
,	O
the	O
process	O
of	O
finding	O
each	O
word	O
in	O
the	O
inverted	B-Algorithm
index	I-Algorithm
(	O
in	O
order	O
to	O
report	O
that	O
it	O
occurred	O
within	O
a	O
document	O
)	O
may	O
be	O
too	O
time	O
consuming	O
,	O
and	O
so	O
this	O
process	O
is	O
commonly	O
split	O
up	O
into	O
two	O
parts	O
,	O
the	O
development	O
of	O
a	O
forward	O
index	O
and	O
a	O
process	O
which	O
sorts	O
the	O
contents	O
of	O
the	O
forward	O
index	O
into	O
the	O
inverted	B-Algorithm
index	I-Algorithm
.	O
</s>
<s>
The	O
inverted	B-Algorithm
index	I-Algorithm
is	O
so	O
named	O
because	O
it	O
is	O
an	O
inversion	O
of	O
the	O
forward	O
index	O
.	O
</s>
<s>
The	O
rationale	O
behind	O
developing	O
a	O
forward	O
index	O
is	O
that	O
as	O
documents	O
are	O
parsed	B-Language
,	O
it	O
is	O
better	O
to	O
intermediately	O
store	O
the	O
words	O
per	O
document	O
.	O
</s>
<s>
The	O
delineation	O
enables	O
asynchronous	O
system	O
processing	O
,	O
which	O
partially	O
circumvents	O
the	O
inverted	B-Algorithm
index	I-Algorithm
update	O
bottleneck	O
.	O
</s>
<s>
The	O
forward	O
index	O
is	O
sorted	B-Algorithm
to	O
transform	O
it	O
to	O
an	O
inverted	B-Algorithm
index	I-Algorithm
.	O
</s>
<s>
Converting	O
the	O
forward	O
index	O
to	O
an	O
inverted	B-Algorithm
index	I-Algorithm
is	O
only	O
a	O
matter	O
of	O
sorting	B-Algorithm
the	O
pairs	O
by	O
the	O
words	O
.	O
</s>
<s>
In	O
this	O
regard	O
,	O
the	O
inverted	B-Algorithm
index	I-Algorithm
is	O
a	O
word-sorted	O
forward	O
index	O
.	O
</s>
<s>
Generating	O
or	O
maintaining	O
a	O
large-scale	O
search	B-Application
engine	I-Application
index	O
represents	O
a	O
significant	O
storage	O
and	O
processing	O
challenge	O
.	O
</s>
<s>
Many	O
search	B-Application
engines	I-Application
utilize	O
a	O
form	O
of	O
compression	B-General_Concept
to	O
reduce	O
the	O
size	O
of	O
the	O
indices	O
on	O
disk	B-General_Concept
.	O
</s>
<s>
Consider	O
the	O
following	O
scenario	O
for	O
a	O
full	B-Application
text	I-Application
,	O
Internet	B-Application
search	I-Application
engine	I-Application
.	O
</s>
<s>
It	O
takes	O
8	O
bits	O
(	O
or	O
1	O
byte	B-Application
)	O
to	O
store	O
a	O
single	O
character	O
.	O
</s>
<s>
Given	O
this	O
scenario	O
,	O
an	O
uncompressed	B-General_Concept
index	O
(	O
assuming	O
a	O
non-conflated	O
,	O
simple	O
,	O
index	O
)	O
for	O
2	O
billion	O
web	O
pages	O
would	O
need	O
to	O
store	O
500	O
billion	O
word	O
entries	O
.	O
</s>
<s>
At	O
1	O
byte	B-Application
per	O
character	O
,	O
or	O
5	O
bytes	B-Application
per	O
word	O
,	O
this	O
would	O
require	O
2500	O
gigabytes	O
of	O
storage	O
space	O
alone	O
.	O
</s>
<s>
Depending	O
on	O
the	O
compression	B-General_Concept
technique	O
chosen	O
,	O
the	O
index	O
can	O
be	O
reduced	O
to	O
a	O
fraction	O
of	O
this	O
size	O
.	O
</s>
<s>
The	O
tradeoff	O
is	O
the	O
time	O
and	O
processing	O
power	O
required	O
to	O
perform	O
compression	B-General_Concept
and	O
decompression	O
.	O
</s>
<s>
Notably	O
,	O
large	O
scale	O
search	B-Application
engine	I-Application
designs	O
incorporate	O
the	O
cost	O
of	O
storage	O
as	O
well	O
as	O
the	O
costs	O
of	O
electricity	O
to	O
power	O
the	O
storage	O
.	O
</s>
<s>
Thus	O
compression	B-General_Concept
is	O
a	O
measure	O
of	O
cost	O
.	O
</s>
<s>
Document	O
parsing	B-Language
breaks	O
apart	O
the	O
components	O
(	O
words	O
)	O
of	O
a	O
document	O
or	O
other	O
form	O
of	O
media	O
for	O
insertion	O
into	O
the	O
forward	O
and	O
inverted	O
indices	O
.	O
</s>
<s>
The	O
words	O
found	O
are	O
called	O
tokens	O
,	O
and	O
so	O
,	O
in	O
the	O
context	O
of	O
search	B-Application
engine	I-Application
indexing	I-Application
and	O
natural	B-Language
language	I-Language
processing	I-Language
,	O
parsing	B-Language
is	O
more	O
commonly	O
referred	O
to	O
as	O
tokenization	O
.	O
</s>
<s>
It	O
is	O
also	O
sometimes	O
called	O
word	O
boundary	O
disambiguation	O
,	O
tagging	O
,	O
text	O
segmentation	O
,	O
content	O
analysis	O
,	O
text	O
analysis	O
,	O
text	B-Algorithm
mining	I-Algorithm
,	O
concordance	O
generation	O
,	O
speech	B-General_Concept
segmentation	I-General_Concept
,	O
lexing	B-Application
,	O
or	O
lexical	B-Application
analysis	I-Application
.	O
</s>
<s>
The	O
terms	O
'	O
indexing	O
 '	O
,	O
'	O
parsing	B-Language
 '	O
,	O
and	O
'	O
tokenization	O
 '	O
are	O
used	O
interchangeably	O
in	O
corporate	O
slang	O
.	O
</s>
<s>
Natural	B-Language
language	I-Language
processing	I-Language
is	O
the	O
subject	O
of	O
continuous	O
research	O
and	O
technological	O
improvement	O
.	O
</s>
<s>
Language-specific	O
logic	O
is	O
employed	O
to	O
properly	O
identify	O
the	O
boundaries	O
of	O
words	O
,	O
which	O
is	O
often	O
the	O
rationale	O
for	O
designing	O
a	O
parser	B-Language
for	O
each	O
language	O
supported	O
(	O
or	O
for	O
groups	O
of	O
languages	O
with	O
similar	O
boundary	O
markers	O
and	O
syntax	O
)	O
.	O
</s>
<s>
Language	O
ambiguity	O
To	O
assist	O
with	O
properly	O
ranking	O
matching	O
documents	O
,	O
many	O
search	B-Application
engines	I-Application
collect	O
additional	O
information	O
about	O
each	O
word	O
,	O
such	O
as	O
its	O
language	O
or	O
lexical	O
category	O
(	O
part	O
of	O
speech	O
)	O
.	O
</s>
<s>
In	O
tokenizing	O
the	O
document	O
,	O
some	O
search	B-Application
engines	I-Application
attempt	O
to	O
automatically	O
identify	O
the	O
language	O
of	O
the	O
document	O
.	O
</s>
<s>
Diverse	O
file	O
formats	O
In	O
order	O
to	O
correctly	O
identify	O
which	O
bytes	B-Application
of	O
a	O
document	O
represent	O
characters	O
,	O
the	O
file	O
format	O
must	O
be	O
correctly	O
handled	O
.	O
</s>
<s>
Search	B-Application
engines	I-Application
which	O
support	O
multiple	O
file	O
formats	O
must	O
be	O
able	O
to	O
correctly	O
open	O
and	O
access	B-General_Concept
the	O
document	O
and	O
be	O
able	O
to	O
tokenize	O
the	O
characters	O
of	O
the	O
document	O
.	O
</s>
<s>
Faulty	O
storage	O
The	O
quality	O
of	O
the	O
natural	B-Language
language	I-Language
data	O
may	O
not	O
always	O
be	O
perfect	O
.	O
</s>
<s>
Unlike	O
literate	O
humans	O
,	O
computers	O
do	O
not	O
understand	O
the	O
structure	O
of	O
a	O
natural	B-Language
language	I-Language
document	O
and	O
cannot	O
automatically	O
recognize	O
words	O
and	O
sentences	O
.	O
</s>
<s>
To	O
a	O
computer	O
,	O
a	O
document	O
is	O
only	O
a	O
sequence	O
of	O
bytes	B-Application
.	O
</s>
<s>
Such	O
a	O
program	O
is	O
commonly	O
called	O
a	O
tokenizer	O
or	O
parser	B-Language
or	O
lexer	B-Application
.	O
</s>
<s>
Many	O
search	B-Application
engines	I-Application
,	O
as	O
well	O
as	O
other	O
natural	B-Language
language	I-Language
processing	I-Language
software	O
,	O
incorporate	O
specialized	B-Application
programs	I-Application
for	O
parsing	B-Language
,	O
such	O
as	O
YACC	B-Application
or	O
Lex	B-General_Concept
.	O
</s>
<s>
During	O
tokenization	O
,	O
the	O
parser	B-Language
identifies	O
sequences	O
of	O
characters	O
which	O
represent	O
words	O
and	O
other	O
elements	O
,	O
such	O
as	O
punctuation	O
,	O
which	O
are	O
represented	O
by	O
numeric	O
codes	O
,	O
some	O
of	O
which	O
are	O
non-printing	O
control	O
characters	O
.	O
</s>
<s>
The	O
parser	B-Language
can	O
also	O
identify	O
entities	B-General_Concept
such	O
as	O
email	O
addresses	O
,	O
phone	O
numbers	O
,	O
and	O
URLs	O
.	O
</s>
<s>
If	O
the	O
search	B-Application
engine	I-Application
supports	O
multiple	O
languages	O
,	O
a	O
common	O
initial	O
step	O
during	O
tokenization	O
is	O
to	O
identify	O
each	O
document	O
's	O
language	O
;	O
many	O
of	O
the	O
subsequent	O
steps	O
are	O
language	O
dependent	O
(	O
such	O
as	O
stemming	B-General_Concept
and	O
part	O
of	O
speech	O
tagging	O
)	O
.	O
</s>
<s>
Language	B-General_Concept
recognition	I-General_Concept
is	O
the	O
process	O
by	O
which	O
a	O
computer	O
program	O
attempts	O
to	O
automatically	O
identify	O
,	O
or	O
categorize	O
,	O
the	O
language	O
of	O
a	O
document	O
.	O
</s>
<s>
Other	O
names	O
for	O
language	B-General_Concept
recognition	I-General_Concept
include	O
language	O
classification	O
,	O
language	O
analysis	O
,	O
language	B-General_Concept
identification	I-General_Concept
,	O
and	O
language	O
tagging	O
.	O
</s>
<s>
Automated	O
language	B-General_Concept
recognition	I-General_Concept
is	O
the	O
subject	O
of	O
ongoing	O
research	O
in	O
natural	B-Language
language	I-Language
processing	I-Language
.	O
</s>
<s>
Finding	O
which	O
language	O
the	O
words	O
belongs	O
to	O
may	O
involve	O
the	O
use	O
of	O
a	O
language	B-General_Concept
recognition	I-General_Concept
chart	O
.	O
</s>
<s>
If	O
the	O
search	B-Application
engine	I-Application
supports	O
multiple	O
document	O
formats	O
,	O
documents	O
must	O
be	O
prepared	O
for	O
tokenization	O
.	O
</s>
<s>
For	O
example	O
,	O
HTML	B-Language
documents	O
contain	O
HTML	B-Language
tags	O
,	O
which	O
specify	O
formatting	O
information	O
such	O
as	O
new	O
line	O
starts	O
,	O
bold	O
emphasis	O
,	O
and	O
font	O
size	O
or	O
style	B-General_Concept
.	O
</s>
<s>
If	O
the	O
search	B-Application
engine	I-Application
were	O
to	O
ignore	O
the	O
difference	O
between	O
content	O
and	O
'	O
markup	O
 '	O
,	O
extraneous	O
information	O
would	O
be	O
included	O
in	O
the	O
index	O
,	O
leading	O
to	O
poor	O
search	B-Application
results	I-Application
.	O
</s>
<s>
Format	O
analysis	O
is	O
also	O
referred	O
to	O
as	O
structure	O
analysis	O
,	O
format	O
parsing	B-Language
,	O
tag	O
stripping	O
,	O
format	O
stripping	O
,	O
text	O
normalization	O
,	O
text	O
cleaning	O
and	O
text	O
preparation	O
.	O
</s>
<s>
Common	O
,	O
well-documented	O
file	O
formats	O
that	O
many	O
search	B-Application
engines	I-Application
support	O
include	O
:	O
</s>
<s>
Options	O
for	O
dealing	O
with	O
various	O
formats	O
include	O
using	O
a	O
publicly	O
available	O
commercial	O
parsing	B-Language
tool	O
that	O
is	O
offered	O
by	O
the	O
organization	O
which	O
developed	O
,	O
maintains	O
,	O
or	O
owns	O
the	O
format	O
,	O
and	O
writing	O
a	O
custom	O
parser	B-Language
.	O
</s>
<s>
Some	O
search	B-Application
engines	I-Application
support	O
inspection	O
of	O
files	O
that	O
are	O
stored	O
in	O
a	O
compressed	B-Operating_System
or	O
encrypted	O
file	O
format	O
.	O
</s>
<s>
When	O
working	O
with	O
a	O
compressed	B-Operating_System
format	O
,	O
the	O
indexer	O
first	O
decompresses	O
the	O
document	O
;	O
this	O
step	O
may	O
result	O
in	O
one	O
or	O
more	O
files	O
,	O
each	O
of	O
which	O
must	O
be	O
indexed	O
separately	O
.	O
</s>
<s>
Commonly	O
supported	O
compressed	B-Operating_System
file	O
formats	O
include	O
:	O
</s>
<s>
hidden	O
"	B-General_Concept
div	I-General_Concept
"	I-General_Concept
tag	I-General_Concept
in	O
HTML	B-Language
,	O
which	O
may	O
incorporate	O
the	O
use	O
of	O
CSS	B-Language
or	O
JavaScript	B-Language
to	O
do	O
so	O
)	O
.	O
</s>
<s>
Some	O
search	B-Application
engines	I-Application
incorporate	O
section	O
recognition	O
,	O
the	O
identification	O
of	O
major	O
parts	O
of	O
a	O
document	O
,	O
prior	O
to	O
tokenization	O
.	O
</s>
<s>
Some	O
file	O
formats	O
,	O
like	O
HTML	B-Language
or	O
PDF	B-Application
,	O
allow	O
for	O
content	O
to	O
be	O
displayed	O
in	O
columns	O
.	O
</s>
<s>
If	O
search	B-Application
engines	I-Application
index	O
this	O
content	O
as	O
if	O
it	O
were	O
normal	O
content	O
,	O
the	O
quality	O
of	O
the	O
index	O
and	O
search	O
quality	O
may	O
be	O
degraded	O
due	O
to	O
the	O
mixed	O
content	O
and	O
improper	O
word	O
proximity	O
.	O
</s>
<s>
Section	O
analysis	O
may	O
require	O
the	O
search	B-Application
engine	I-Application
to	O
implement	O
the	O
rendering	O
logic	O
of	O
each	O
document	O
,	O
essentially	O
an	O
abstract	O
representation	O
of	O
the	O
actual	O
document	O
,	O
and	O
then	O
index	O
the	O
representation	O
instead	O
.	O
</s>
<s>
For	O
example	O
,	O
some	O
content	O
on	O
the	O
Internet	O
is	O
rendered	O
via	O
JavaScript	B-Language
.	O
</s>
<s>
If	O
the	O
search	B-Application
engine	I-Application
does	O
not	O
render	O
the	O
page	O
and	O
evaluate	O
the	O
JavaScript	B-Language
within	O
the	O
page	O
,	O
it	O
would	O
not	O
'	O
see	O
 '	O
this	O
content	O
in	O
the	O
same	O
way	O
and	O
would	O
index	O
the	O
document	O
incorrectly	O
.	O
</s>
<s>
Given	O
that	O
some	O
search	B-Application
engines	I-Application
do	O
not	O
bother	O
with	O
rendering	O
issues	O
,	O
many	O
web	O
page	O
designers	O
avoid	O
displaying	O
content	O
via	O
JavaScript	B-Language
or	O
use	O
the	O
tag	O
to	O
ensure	O
that	O
the	O
web	O
page	O
is	O
indexed	O
properly	O
.	O
</s>
<s>
At	O
the	O
same	O
time	O
,	O
this	O
fact	O
can	O
also	O
be	O
exploited	O
to	O
cause	O
the	O
search	B-Application
engine	I-Application
indexer	O
to	O
'	O
see	O
 '	O
different	O
content	O
than	O
the	O
viewer	O
.	O
</s>
<s>
Indexing	O
often	O
has	O
to	O
recognize	O
the	O
HTML	B-Language
tags	O
to	O
organize	O
priority	O
.	O
</s>
<s>
Some	O
indexers	O
like	O
Google	B-Application
and	O
Bing	B-Application
ensure	O
that	O
the	O
search	B-Application
engine	I-Application
does	O
not	O
take	O
the	O
large	O
texts	O
as	O
relevant	O
source	O
due	O
to	O
strong	O
type	O
system	O
compatibility	O
.	O
</s>
<s>
For	O
HTML	B-Language
pages	O
,	O
the	O
meta	O
tag	O
contains	O
keywords	O
which	O
are	O
also	O
included	O
in	O
the	O
index	O
.	O
</s>
<s>
Earlier	O
Internet	B-Application
search	I-Application
engine	I-Application
technology	O
would	O
only	O
index	O
the	B-Application
keywords	I-Application
in	O
the	O
meta	O
tags	O
for	O
the	O
forward	O
index	O
;	O
the	O
full	O
document	O
would	O
not	O
be	O
parsed	B-Language
.	O
</s>
<s>
At	O
that	O
time	O
full-text	B-Application
indexing	O
was	O
not	O
as	O
well	O
established	O
,	O
nor	O
was	O
computer	B-Architecture
hardware	I-Architecture
able	O
to	O
support	O
such	O
technology	O
.	O
</s>
<s>
The	O
design	O
of	O
the	O
HTML	B-Language
markup	O
language	O
initially	O
included	O
support	O
for	O
meta	O
tags	O
for	O
the	O
very	O
purpose	O
of	O
being	O
properly	O
and	O
easily	O
indexed	O
,	O
without	O
requiring	O
tokenization	O
.	O
</s>
<s>
The	B-Application
keywords	I-Application
used	O
to	O
describe	O
webpages	O
(	O
many	O
of	O
which	O
were	O
corporate-oriented	O
webpages	O
similar	O
to	O
product	O
brochures	O
)	O
changed	O
from	O
descriptive	O
to	O
marketing-oriented	O
keywords	O
designed	O
to	O
drive	O
sales	O
by	O
placing	O
the	O
webpage	O
high	O
in	O
the	O
search	B-Application
results	I-Application
for	O
specific	O
search	O
queries	B-Library
.	O
</s>
<s>
The	O
fact	O
that	O
these	O
keywords	O
were	O
subjectively	O
specified	O
was	O
leading	O
to	O
spamdexing	O
,	O
which	O
drove	O
many	O
search	B-Application
engines	I-Application
to	O
adopt	O
full-text	B-Application
indexing	O
technologies	O
in	O
the	O
1990s	O
.	O
</s>
<s>
Search	B-Application
engine	I-Application
designers	O
and	O
companies	O
could	O
only	O
place	O
so	O
many	O
'	O
marketing	O
keywords	O
 '	O
into	O
the	O
content	O
of	O
a	O
webpage	O
before	O
draining	O
it	O
of	O
all	O
interesting	O
and	O
useful	O
information	O
.	O
</s>
<s>
In	O
this	O
sense	O
,	O
full-text	B-Application
indexing	O
was	O
more	O
objective	O
and	O
increased	O
the	O
quality	O
of	O
search	B-Application
engine	I-Application
results	O
,	O
as	O
it	O
was	O
one	O
more	O
step	O
away	O
from	O
subjective	O
control	O
of	O
search	B-Application
engine	I-Application
result	O
placement	O
,	O
which	O
in	O
turn	O
furthered	O
research	O
of	O
full-text	B-Application
indexing	O
technologies	O
.	O
</s>
<s>
In	O
desktop	B-Application
search	I-Application
,	O
many	O
solutions	O
incorporate	O
meta	O
tags	O
to	O
provide	O
a	O
way	O
for	O
authors	O
to	O
further	O
customize	O
how	O
the	O
search	B-Application
engine	I-Application
will	O
index	O
content	O
from	O
various	O
files	O
that	O
is	O
not	O
evident	O
from	O
the	O
file	O
content	O
.	O
</s>
<s>
Desktop	B-Application
search	I-Application
is	O
more	O
under	O
the	O
control	O
of	O
the	O
user	O
,	O
while	O
Internet	B-Application
search	I-Application
engines	I-Application
must	O
focus	O
more	O
on	O
the	O
full	B-Application
text	I-Application
index	I-Application
.	O
</s>
