<s>
Unstructured	B-Application
data	I-Application
(	O
or	O
unstructured	B-Application
information	I-Application
)	O
is	O
information	O
that	O
either	O
does	O
not	O
have	O
a	O
pre-defined	O
data	B-Application
model	I-Application
or	O
is	O
not	O
organized	O
in	O
a	O
pre-defined	O
manner	O
.	O
</s>
<s>
Unstructured	B-Application
information	I-Application
is	O
typically	O
text-heavy	O
,	O
but	O
may	O
contain	O
data	O
such	O
as	O
dates	O
,	O
numbers	O
,	O
and	O
facts	O
as	O
well	O
.	O
</s>
<s>
This	O
results	O
in	O
irregularities	O
and	O
ambiguities	O
that	O
make	O
it	O
difficult	O
to	O
understand	O
using	O
traditional	O
programs	O
as	O
compared	O
to	O
data	O
stored	O
in	O
fielded	O
form	O
in	O
databases	O
or	O
annotated	O
(	O
semantically	B-General_Concept
tagged	I-General_Concept
)	O
in	O
documents	O
.	O
</s>
<s>
In	O
1998	O
,	O
Merrill	O
Lynch	O
said	O
"	O
unstructured	B-Application
data	I-Application
comprises	O
the	O
vast	O
majority	O
of	O
data	O
found	O
in	O
an	O
organization	O
,	O
some	O
estimates	O
run	O
as	O
high	O
as	O
80%	O
.	O
"	O
</s>
<s>
Other	O
sources	O
have	O
reported	O
similar	O
or	O
higher	O
percentages	O
of	O
unstructured	B-Application
data	I-Application
.	O
</s>
<s>
More	O
recently	O
,	O
IDC	O
and	O
Seagate	O
predict	O
that	O
the	O
global	O
datasphere	B-General_Concept
will	O
grow	O
to	O
163	O
zettabytes	O
by	O
2025	O
and	O
majority	O
of	O
that	O
will	O
be	O
unstructured	O
.	O
</s>
<s>
The	O
Computer	O
World	O
magazine	O
states	O
that	O
unstructured	B-Application
information	I-Application
might	O
account	O
for	O
more	O
than	O
70	O
–	O
80%	O
of	O
all	O
data	O
in	O
organizations	O
.	O
</s>
<s>
The	O
earliest	O
research	O
into	O
business	B-General_Concept
intelligence	I-General_Concept
focused	O
in	O
on	O
unstructured	O
textual	O
data	O
,	O
rather	O
than	O
numerical	O
data	O
.	O
</s>
<s>
As	O
early	O
as	O
1958	O
,	O
computer	B-General_Concept
science	I-General_Concept
researchers	O
like	O
H.P.	O
</s>
<s>
In	O
2004	O
,	O
the	O
SAS	B-Language
Institute	O
developed	O
the	O
SAS	B-Language
Text	O
Miner	O
,	O
which	O
uses	O
Singular	O
Value	O
Decomposition	O
(	O
SVD	O
)	O
to	O
reduce	O
a	O
hyper-dimensional	O
textual	O
space	O
into	O
smaller	O
dimensions	O
for	O
significantly	O
more	O
efficient	O
machine-analysis	O
.	O
</s>
<s>
The	O
emergence	O
of	O
Big	B-Application
Data	I-Application
in	O
the	O
late	O
2000s	O
led	O
to	O
a	O
heightened	O
interest	O
in	O
the	O
applications	O
of	O
unstructured	B-Application
data	I-Application
analytics	O
in	O
contemporary	O
fields	O
such	O
as	O
predictive	B-General_Concept
analytics	I-General_Concept
and	O
root	O
cause	O
analysis	O
.	O
</s>
<s>
Unstructured	B-Application
information	I-Application
might	O
have	O
some	O
structure	O
(	O
semi-structured	B-Application
)	O
or	O
even	O
be	O
highly	O
structured	O
but	O
in	O
ways	O
that	O
are	O
unanticipated	O
or	O
unannounced	O
.	O
</s>
<s>
Techniques	O
such	O
as	O
data	B-Application
mining	I-Application
,	O
natural	B-Language
language	I-Language
processing	I-Language
(	O
NLP	B-Language
)	O
,	O
and	O
text	B-Algorithm
analytics	I-Algorithm
provide	O
different	O
methods	O
to	O
find	O
patterns	O
in	O
,	O
or	O
otherwise	O
interpret	O
,	O
this	O
information	O
.	O
</s>
<s>
Common	O
techniques	O
for	O
structuring	O
text	O
usually	O
involve	O
manual	O
tagging	B-General_Concept
with	I-General_Concept
metadata	I-General_Concept
or	O
part-of-speech	O
tagging	O
for	O
further	O
text	O
mining-based	O
structuring	O
.	O
</s>
<s>
The	O
Unstructured	B-Application
Information	I-Application
Management	I-Application
Architecture	I-Application
(	O
UIMA	B-Application
)	O
standard	O
provided	O
a	O
common	O
framework	O
for	O
processing	O
this	O
information	O
to	O
extract	O
meaning	O
and	O
create	O
structured	B-Application
data	I-Application
about	O
the	O
information	O
.	O
</s>
<s>
Unstructured	B-Application
information	I-Application
can	O
then	O
be	O
enriched	O
and	O
tagged	O
to	O
address	O
ambiguities	O
and	O
relevancy-based	O
techniques	O
then	O
used	O
to	O
facilitate	O
search	O
and	O
discovery	O
.	O
</s>
<s>
Examples	O
of	O
"	O
unstructured	B-Application
data	I-Application
"	O
may	O
include	O
books	O
,	O
journals	O
,	O
documents	O
,	O
metadata	O
,	O
health	O
records	O
,	O
audio	O
,	O
video	O
,	O
analog	O
data	O
,	O
images	O
,	O
files	O
,	O
and	O
unstructured	O
text	O
such	O
as	O
the	O
body	O
of	O
an	O
e-mail	O
message	O
,	O
Web	O
page	O
,	O
or	O
word-processor	B-General_Concept
document	O
.	O
</s>
<s>
While	O
the	O
main	O
content	B-Protocol
being	O
conveyed	O
does	O
not	O
have	O
a	O
defined	O
structure	O
,	O
it	O
generally	O
comes	O
packaged	O
in	O
objects	O
(	O
e.g.	O
</s>
<s>
in	O
files	O
or	O
documents	O
,	O
...	O
)	O
that	O
themselves	O
have	O
structure	O
and	O
are	O
thus	O
a	O
mix	O
of	O
structured	O
and	O
unstructured	B-Application
data	I-Application
,	O
but	O
collectively	O
this	O
is	O
still	O
referred	O
to	O
as	O
"	O
unstructured	B-Application
data	I-Application
"	O
.	O
</s>
<s>
For	O
example	O
,	O
an	O
HTML	B-Language
web	O
page	O
is	O
tagged	O
,	O
but	O
HTML	B-Language
mark-up	O
typically	O
serves	O
solely	O
for	O
rendering	O
.	O
</s>
<s>
It	O
does	O
not	O
capture	O
the	O
meaning	O
or	O
function	O
of	O
tagged	O
elements	O
in	O
ways	O
that	O
support	O
automated	O
processing	O
of	O
the	O
information	O
content	B-Protocol
of	O
the	O
page	O
.	O
</s>
<s>
XHTML	B-Language
tagging	O
does	O
allow	O
machine	O
processing	O
of	O
elements	O
,	O
although	O
it	O
typically	O
does	O
not	O
capture	O
or	O
convey	O
the	O
semantic	O
meaning	O
of	O
tagged	O
terms	O
.	O
</s>
<s>
Since	O
unstructured	B-Application
data	I-Application
commonly	O
occurs	O
in	O
electronic	B-Protocol
documents	I-Protocol
,	O
the	O
use	O
of	O
a	O
content	B-Protocol
or	O
document	O
management	O
system	O
which	O
can	O
categorize	O
entire	O
documents	O
is	O
often	O
preferred	O
over	O
data	O
transfer	O
and	O
manipulation	O
from	O
within	O
the	O
documents	O
.	O
</s>
<s>
Search	B-Application
engines	I-Application
have	O
become	O
popular	O
tools	O
for	O
indexing	O
and	O
searching	O
through	O
such	O
data	O
,	O
especially	O
text	O
.	O
</s>
<s>
Specific	O
computational	O
workflows	O
have	O
been	O
developed	O
to	O
impose	O
structure	O
upon	O
the	O
unstructured	B-Application
data	I-Application
contained	O
within	O
text	O
documents	O
.	O
</s>
<s>
Several	O
of	O
these	O
approaches	O
are	O
based	O
upon	O
the	O
concept	O
of	O
online	B-Application
analytical	I-Application
processing	I-Application
,	I-Application
or	I-Application
OLAP	I-Application
,	O
and	O
may	O
be	O
supported	O
by	O
data	B-Application
models	I-Application
such	O
as	O
text	O
cubes	O
.	O
</s>
<s>
Once	O
document	O
metadata	O
is	O
available	O
through	O
a	O
data	B-Application
model	I-Application
,	O
generating	O
summaries	O
of	O
subsets	O
of	O
documents	O
(	O
i.e.	O
,	O
cells	O
within	O
a	O
text	O
cube	O
)	O
may	O
be	O
performed	O
with	O
phrase-based	O
approaches	O
.	O
</s>
<s>
Biomedical	O
research	O
generates	O
one	O
major	O
source	O
of	O
unstructured	B-Application
data	I-Application
as	O
researchers	O
often	O
publish	O
their	O
findings	O
in	O
scholarly	O
journals	O
.	O
</s>
<s>
Recent	O
efforts	O
to	O
enforce	O
structure	O
upon	O
biomedical	O
documents	O
include	O
self-organizing	B-Algorithm
map	I-Algorithm
approaches	O
for	O
identifying	O
topics	O
among	O
documents	O
,	O
general-purpose	O
unsupervised	B-General_Concept
algorithms	I-General_Concept
,	O
and	O
an	O
application	O
of	O
the	O
CaseOLAP	O
workflow	O
to	O
determine	O
associations	O
between	O
protein	O
names	O
and	O
cardiovascular	O
disease	O
topics	O
in	O
the	O
literature	O
.	O
</s>
<s>
This	O
terminology	O
,	O
unstructured	B-Application
data	I-Application
,	O
is	O
rarely	O
used	O
in	O
the	O
EU	O
after	O
GDPR	O
came	O
into	O
force	O
in	O
2018	O
.	O
</s>
<s>
GDPR	O
does	O
neither	O
mention	O
nor	O
define	O
"	O
unstructured	B-Application
data	I-Application
"	O
.	O
</s>
