<s>
Speaker	B-Application
diarisation	I-Application
(	O
or	O
diarization	B-Application
)	O
is	O
the	O
process	O
of	O
partitioning	O
an	O
audio	O
stream	O
containing	O
human	O
speech	O
into	O
homogeneous	O
segments	O
according	O
to	O
the	O
identity	O
of	O
each	O
speaker	O
.	O
</s>
<s>
It	O
can	O
enhance	O
the	O
readability	O
of	O
an	O
automatic	B-Application
speech	I-Application
transcription	I-Application
by	O
structuring	O
the	O
audio	O
stream	O
into	O
speaker	O
turns	O
and	O
,	O
when	O
used	O
together	O
with	O
speaker	B-Application
recognition	I-Application
systems	O
,	O
by	O
providing	O
the	O
speaker	O
’s	O
true	O
identity	O
.	O
</s>
<s>
Speaker	B-Application
diarisation	I-Application
is	O
a	O
combination	O
of	O
speaker	O
segmentation	O
and	O
speaker	O
clustering	O
.	O
</s>
<s>
With	O
the	O
increasing	O
number	O
of	O
broadcasts	O
,	O
meeting	O
recordings	O
and	O
voice	O
mail	O
collected	O
every	O
year	O
,	O
speaker	B-Application
diarisation	I-Application
has	O
received	O
much	O
attention	O
by	O
the	O
speech	O
community	O
,	O
as	O
is	O
manifested	O
by	O
the	O
specific	O
evaluations	O
devoted	O
to	O
it	O
under	O
the	O
auspices	O
of	O
the	O
National	O
Institute	O
of	O
Standards	O
and	O
Technology	O
for	O
telephone	O
speech	O
,	O
broadcast	O
news	O
and	O
meetings	O
.	O
</s>
<s>
In	O
speaker	B-Application
diarisation	I-Application
,	O
one	O
of	O
the	O
most	O
popular	O
methods	O
is	O
to	O
use	O
a	O
Gaussian	O
mixture	O
model	O
to	O
model	O
each	O
of	O
the	O
speakers	O
,	O
and	O
assign	O
the	O
corresponding	O
frames	O
for	O
each	O
speaker	O
with	O
the	O
help	O
of	O
a	O
Hidden	O
Markov	O
Model	O
.	O
</s>
<s>
More	O
recently	O
,	O
speaker	B-Application
diarisation	I-Application
is	O
performed	O
thanks	O
to	O
neural	B-Architecture
networks	I-Architecture
and	O
heavier	O
GPU	B-Architecture
computing	O
made	O
possible	O
some	O
more	O
efficient	O
diarisation	O
algorithm	O
.	O
</s>
<s>
There	O
are	O
some	O
open	O
source	O
initiatives	O
for	O
speaker	B-Application
diarisation	I-Application
(	O
in	O
alphabetical	O
order	O
)	O
:	O
</s>
<s>
ALIZE	O
Speaker	B-Application
Diarization	I-Application
(	O
last	O
repository	O
update	O
:	O
July	O
2016	O
;	O
last	O
release	O
:	O
February	O
2013	O
,	O
version	O
:	O
3.0	O
)	O
:	O
ALIZE	O
Diarization	B-Application
System	O
,	O
developed	O
at	O
the	O
University	O
Of	O
Avignon	O
,	O
a	O
release	O
2.0	O
is	O
available	O
.	O
</s>
<s>
pyannote.audio	O
(	O
last	O
repository	O
update	O
:	O
August	O
2022	O
,	O
last	O
release	O
:	O
July	O
2022	O
,	O
version	O
:	O
2.0	O
)	O
:	O
pyannote.audio	O
is	O
an	O
open-source	O
toolkit	O
written	O
in	O
Python	O
for	O
speaker	B-Application
diarization	I-Application
.	O
</s>
<s>
SHoUT	O
(	O
last	O
update	O
:	O
December	O
2010	O
;	O
version	O
:	O
0.3	O
)	O
:	O
SHoUT	O
is	O
a	O
software	O
package	O
developed	O
at	O
the	O
University	O
of	O
Twente	O
to	O
aid	O
speech	B-Application
recognition	I-Application
research	O
.	O
</s>
<s>
SHoUT	O
is	O
a	O
Dutch	O
acronym	O
for	O
Speech	B-Application
Recognition	I-Application
Research	O
at	O
the	O
University	O
of	O
Twente	O
.	O
</s>
