<s>
CMU	B-General_Concept
Sphinx	I-General_Concept
,	O
also	O
called	O
Sphinx	O
for	O
short	O
,	O
is	O
the	O
general	O
term	O
to	O
describe	O
a	O
group	O
of	O
speech	B-Application
recognition	I-Application
systems	O
developed	O
at	O
Carnegie	O
Mellon	O
University	O
.	O
</s>
<s>
These	O
include	O
a	O
series	O
of	O
speech	B-Application
recognizers	I-Application
(	O
Sphinx	O
2	O
-	O
4	O
)	O
and	O
an	O
acoustic	B-General_Concept
model	I-General_Concept
trainer	O
(	O
SphinxTrain	O
)	O
.	O
</s>
<s>
In	O
2000	O
,	O
the	O
Sphinx	O
group	O
at	O
Carnegie	O
Mellon	O
committed	O
to	O
open	O
source	O
several	O
speech	B-Application
recognizer	I-Application
components	O
,	O
including	O
Sphinx	O
2	O
and	O
later	O
Sphinx	O
3	O
(	O
in	O
2001	O
)	O
.	O
</s>
<s>
The	O
speech	O
decoders	O
come	O
with	O
acoustic	B-General_Concept
models	I-General_Concept
and	O
sample	O
applications	O
.	O
</s>
<s>
The	O
available	O
resources	O
include	O
in	O
addition	O
software	O
for	O
acoustic	B-General_Concept
model	I-General_Concept
training	O
,	O
language	B-Language
model	I-Language
compilation	O
and	O
a	O
public	O
domain	O
pronunciation	O
dictionary	O
,	O
cmudict	B-General_Concept
.	O
</s>
<s>
Sphinx	O
is	O
a	O
continuous-speech	O
,	O
speaker-independent	O
recognition	O
system	O
making	O
use	O
of	O
hidden	O
Markov	O
acoustic	B-General_Concept
models	I-General_Concept
(	O
HMMs	O
)	O
and	O
an	O
n-gram	B-Language
statistical	B-Language
language	I-Language
model	I-Language
.	O
</s>
<s>
A	O
fast	O
performance-oriented	O
recognizer	O
,	O
originally	O
developed	O
by	O
Xuedong	O
Huang	O
at	O
Carnegie	O
Mellon	O
and	O
released	O
as	O
open-source	B-Application
with	O
a	O
BSD-style	B-Operating_System
license	I-Operating_System
on	O
SourceForge	B-Application
by	O
Kevin	O
Lenzo	O
at	O
LinuxWorld	O
in	O
2000	O
.	O
</s>
<s>
As	O
such	O
it	O
incorporates	O
functionality	O
such	O
as	O
end-pointing	O
,	O
partial	O
hypothesis	O
generation	O
,	O
dynamic	O
language	B-Language
model	I-Language
switching	O
and	O
so	O
on	O
.	O
</s>
<s>
It	O
can	O
be	O
used	O
in	O
computer	O
based	O
PBX	O
systems	O
such	O
as	O
Asterisk	B-Language
.	O
</s>
<s>
Sphinx	O
3	O
is	O
under	O
active	O
development	O
and	O
in	O
conjunction	O
with	O
SphinxTrain	O
provides	O
access	O
to	O
a	O
number	O
of	O
modern	O
modeling	O
techniques	O
,	O
such	O
as	O
LDA/MLLT	O
,	O
MLLR	O
and	O
VTLN	O
,	O
that	O
improve	O
recognition	O
accuracy	O
(	O
see	O
the	O
article	O
on	O
Speech	B-Application
Recognition	I-Application
for	O
descriptions	O
of	O
these	O
techniques	O
)	O
.	O
</s>
<s>
Sphinx	O
4	O
is	O
a	O
complete	O
rewrite	O
of	O
the	O
Sphinx	O
engine	O
with	O
the	O
goal	O
of	O
providing	O
a	O
more	O
flexible	O
framework	O
for	O
research	O
in	O
speech	B-Application
recognition	I-Application
,	O
written	O
entirely	O
in	O
the	O
Java	B-Language
programming	I-Language
language	I-Language
.	O
</s>
<s>
(	O
Currently	O
supported	O
languages	O
are	O
C	B-Language
,	O
C++	O
,	O
C#	O
,	O
Python	O
,	O
Ruby	O
,	O
Java	B-Language
,	O
and	O
JavaScript	O
.	O
)	O
</s>
<s>
A	O
version	O
of	O
Sphinx	O
that	O
can	O
be	O
used	O
in	O
embedded	O
systems	O
(	O
e.g.	O
,	O
based	O
on	O
an	O
ARM	B-Architecture
processor	I-Architecture
)	O
.	O
</s>
