<s>
a	O
textual	O
document	O
(	O
or	O
a	O
set	O
of	O
documents	O
)	O
by	O
reducing	O
language	O
heterogeneity	O
,	O
while	O
maintaining	O
text	O
semantics	B-Application
.	O
</s>
<s>
In	O
most	O
applications	O
,	O
semantic	B-General_Concept
compression	I-General_Concept
is	O
a	O
lossy	O
compression	O
,	O
that	O
is	O
,	O
increased	O
prolixity	O
does	O
not	O
compensate	O
for	O
the	O
lexical	O
compression	O
,	O
and	O
an	O
original	O
document	O
cannot	O
be	O
reconstructed	O
in	O
a	O
reverse	O
process	O
.	O
</s>
<s>
Semantic	B-General_Concept
compression	I-General_Concept
is	O
basically	O
achieved	O
in	O
two	O
steps	O
,	O
using	O
frequency	O
dictionaries	O
and	O
semantic	O
network	O
:	O
</s>
<s>
replacing	O
less	O
frequent	O
terms	O
with	O
their	O
hypernyms	O
(	O
generalization	B-Algorithm
)	O
from	O
target	O
lexicon	O
.	O
</s>
<s>
The	O
below	O
fragment	O
of	O
text	O
has	O
been	O
processed	O
by	O
the	O
semantic	B-General_Concept
compression	I-General_Concept
.	O
</s>
<s>
A	O
natural	O
tendency	O
to	O
keep	O
natural	O
language	O
expressions	O
concise	O
can	O
be	O
perceived	O
as	O
a	O
form	O
of	O
implicit	O
semantic	B-General_Concept
compression	I-General_Concept
,	O
by	O
omitting	O
unmeaningful	O
words	O
or	O
redundant	O
meaningful	O
words	O
(	O
especially	O
to	O
avoid	O
pleonasms	O
)	O
.	O
</s>
<s>
Semantic	B-General_Concept
compression	I-General_Concept
is	O
advantageous	O
in	O
information	B-Library
retrieval	I-Library
tasks	O
,	O
improving	O
their	O
effectiveness	O
(	O
in	O
terms	O
of	O
both	O
precision	O
and	O
recall	O
)	O
.	O
</s>
