<s>
Ontology	B-General_Concept
learning	I-General_Concept
(	O
ontology	B-General_Concept
extraction	I-General_Concept
,	O
ontology	B-General_Concept
generation	I-General_Concept
,	O
or	O
ontology	B-General_Concept
acquisition	I-General_Concept
)	O
is	O
the	O
automatic	O
or	O
semi-automatic	O
creation	O
of	O
ontologies	B-Language
,	O
including	O
extracting	O
the	O
corresponding	O
domain	B-Algorithm
's	I-Algorithm
terms	O
and	O
the	O
relationships	O
between	O
the	O
concepts	O
that	O
these	O
terms	O
represent	O
from	O
a	O
corpus	O
of	O
natural	O
language	O
text	O
,	O
and	O
encoding	O
them	O
with	O
an	O
ontology	B-Language
language	I-Language
for	O
easy	O
retrieval	O
.	O
</s>
<s>
As	O
building	O
ontologies	B-Language
manually	O
is	O
extremely	O
labor-intensive	O
and	O
time-consuming	O
,	O
there	O
is	O
great	O
motivation	O
to	O
automate	O
the	O
process	O
.	O
</s>
<s>
Ontology	B-General_Concept
learning	I-General_Concept
(	O
OL	O
)	O
is	O
used	O
to	O
(	O
semi	O
-	O
)	O
automatically	O
extract	O
whole	O
ontologies	B-Language
from	O
natural	O
language	O
text	O
.	O
</s>
<s>
The	O
process	O
is	O
usually	O
split	O
into	O
the	O
following	O
eight	O
tasks	O
,	O
which	O
are	O
not	O
all	O
necessarily	O
applied	O
in	O
every	O
ontology	B-General_Concept
learning	I-General_Concept
system	O
.	O
</s>
<s>
In	O
the	O
subsequent	O
step	O
,	O
similarly	O
to	O
coreference	O
resolution	O
in	O
information	B-General_Concept
extraction	I-General_Concept
,	O
the	O
OL	O
system	O
determines	O
synonyms	O
,	O
because	O
they	O
share	O
the	O
same	O
meaning	O
and	O
therefore	O
correspond	O
to	O
the	O
same	O
concept	O
.	O
</s>
<s>
This	O
is	O
mostly	O
achieved	O
with	O
unsupervised	O
hierarchical	B-Algorithm
clustering	I-Algorithm
methods	O
.	O
</s>
<s>
Instead	O
,	O
bootstrapping	B-Application
methods	I-Application
are	O
developed	O
,	O
which	O
learn	O
these	O
patterns	O
automatically	O
and	O
therefore	O
ensure	O
broader	O
coverage	O
.	O
</s>
<s>
During	O
rule	B-Algorithm
discovery	I-Algorithm
,	O
axioms	O
(	O
formal	O
description	O
of	O
concepts	O
)	O
are	O
generated	O
for	O
the	O
extracted	O
concepts	O
.	O
</s>
<s>
At	O
this	O
step	O
,	O
the	O
ontology	B-Language
is	O
augmented	O
with	O
instances	O
of	O
concepts	O
and	O
properties	O
.	O
</s>
<s>
Instances	O
of	O
properties	O
are	O
added	O
through	O
the	O
application	O
of	O
bootstrapping	B-Application
methods	I-Application
,	O
which	O
collect	O
relation	O
tuples	O
.	O
</s>
<s>
In	O
this	O
step	O
,	O
the	O
OL	O
system	O
tries	O
to	O
extend	O
the	O
taxonomic	O
structure	O
of	O
an	O
existing	O
ontology	B-Language
with	O
further	O
concepts	O
.	O
</s>
<s>
Approaches	O
range	O
from	O
applying	O
SVM	O
with	O
kernel	B-Algorithm
methods	I-Algorithm
to	O
semantic	O
role	O
labeling	O
(	O
SRL	O
)	O
to	O
deep	O
semantic	B-Application
parsing	I-Application
techniques	O
.	O
</s>
<s>
Dog4Dag	O
(	O
Dresden	O
Ontology	B-Language
Generator	O
for	O
Directed	O
Acyclic	O
Graphs	O
)	O
is	O
an	O
ontology	B-General_Concept
generation	I-General_Concept
plugin	O
for	O
Protégé	O
4.1	O
and	O
OBOEdit	O
2.1	O
.	O
</s>
<s>
Integrated	O
into	O
Protégé	O
4.1	O
and	O
OBO-Edit	O
2.1	O
,	O
DOG4DAG	O
allows	O
ontology	B-Language
extension	O
for	O
all	O
common	O
ontology	B-Language
formats	O
(	O
e.g.	O
,	O
OWL	O
and	O
OBO	O
)	O
.	O
</s>
