<s>
Speech	B-General_Concept
segmentation	I-General_Concept
is	O
the	O
process	O
of	O
identifying	O
the	O
boundaries	O
between	O
words	O
,	O
syllables	B-General_Concept
,	O
or	O
phonemes	B-Language
in	O
spoken	O
natural	O
languages	O
.	O
</s>
<s>
The	O
term	O
applies	O
both	O
to	O
the	O
mental	O
processes	O
used	O
by	O
humans	O
,	O
and	O
to	O
artificial	O
processes	O
of	O
natural	B-Language
language	I-Language
processing	I-Language
.	O
</s>
<s>
Speech	B-General_Concept
segmentation	I-General_Concept
is	O
a	O
subfield	O
of	O
general	O
speech	O
perception	O
and	O
an	O
important	O
subproblem	O
of	O
the	O
technologically	O
focused	O
field	O
of	O
speech	B-Application
recognition	I-Application
,	O
and	O
cannot	O
be	O
adequately	O
solved	O
in	O
isolation	O
.	O
</s>
<s>
As	O
in	O
most	O
natural	B-Language
language	I-Language
processing	I-Language
problems	O
,	O
one	O
must	O
take	O
into	O
account	O
context	O
,	O
grammar	O
,	O
and	O
semantics	B-Application
,	O
and	O
even	O
so	O
the	O
result	O
is	O
often	O
a	O
probabilistic	O
division	O
(	O
statistically	O
based	O
on	O
likelihood	O
)	O
rather	O
than	O
a	O
categorical	O
one	O
.	O
</s>
<s>
Though	O
it	O
seems	O
that	O
coarticulation	O
—	O
a	O
phenomenon	O
which	O
may	O
happen	O
between	O
adjacent	O
words	O
just	O
as	O
easily	O
as	O
within	O
a	O
single	O
word	O
—	O
presents	O
the	O
main	O
challenge	O
in	O
speech	B-General_Concept
segmentation	I-General_Concept
across	O
languages	O
,	O
some	O
other	O
problems	O
and	O
strategies	O
employed	O
in	O
solving	O
those	O
problems	O
can	O
be	O
seen	O
in	O
the	O
following	O
sections	O
.	O
</s>
<s>
This	O
problem	O
overlaps	O
to	O
some	O
extent	O
with	O
the	O
problem	O
of	O
text	O
segmentation	O
that	O
occurs	O
in	O
some	O
languages	O
which	O
are	O
traditionally	O
written	O
without	O
inter-word	O
spaces	O
,	O
like	O
Chinese	O
and	O
Japanese	O
,	O
compared	O
to	O
writing	O
systems	O
which	O
indicate	O
speech	B-General_Concept
segmentation	I-General_Concept
between	O
words	O
by	O
a	O
word	O
divider	O
,	O
such	O
as	O
the	O
space	O
.	O
</s>
<s>
However	O
,	O
even	O
for	O
those	O
languages	O
,	O
text	O
segmentation	O
is	O
often	O
much	O
easier	O
than	O
speech	B-General_Concept
segmentation	I-General_Concept
,	O
because	O
the	O
written	O
language	O
usually	O
has	O
little	O
interference	O
between	O
adjacent	O
words	O
,	O
and	O
often	O
contains	O
additional	O
clues	O
not	O
present	O
in	O
speech	O
(	O
such	O
as	O
the	O
use	O
of	O
Chinese	O
characters	O
for	O
word	O
stems	O
in	O
Japanese	O
)	O
.	O
</s>
<s>
Though	O
lexical	O
recognition	O
is	O
not	O
thought	O
to	O
be	O
used	O
by	O
infants	O
in	O
their	O
first	O
year	O
,	O
due	O
to	O
their	O
highly	O
limited	O
vocabularies	O
,	O
it	O
is	O
one	O
of	O
the	O
major	O
processes	O
involved	O
in	O
speech	B-General_Concept
segmentation	I-General_Concept
for	O
adults	O
.	O
</s>
<s>
In	O
any	O
case	O
,	O
lexical	O
recognition	O
likely	O
contributes	O
significantly	O
to	O
speech	B-General_Concept
segmentation	I-General_Concept
through	O
the	O
contextual	O
clues	O
it	O
provides	O
,	O
given	O
that	O
it	O
is	O
a	O
heavily	O
probabilistic	O
system	O
—	O
based	O
on	O
the	O
statistical	O
likelihood	O
of	O
certain	O
words	O
or	O
constituents	O
occurring	O
together	O
.	O
</s>
<s>
As	O
this	O
example	O
shows	O
,	O
proper	O
lexical	O
segmentation	O
depends	O
on	O
context	O
and	O
semantics	B-Application
which	O
draws	O
on	O
the	O
whole	O
of	O
human	O
knowledge	O
and	O
experience	O
,	O
and	O
would	O
thus	O
require	O
advanced	O
pattern	O
recognition	O
and	O
artificial	B-Application
intelligence	I-Application
technologies	O
to	O
be	O
implemented	O
on	O
a	O
computer	O
.	O
</s>
<s>
Lexical	O
recognition	O
is	O
of	O
particular	O
value	O
in	O
the	O
field	O
of	O
computer	B-Application
speech	I-Application
recognition	I-Application
,	O
since	O
the	O
ability	O
to	O
build	O
and	O
search	O
a	O
network	O
of	O
semantically	O
connected	O
ideas	O
would	O
greatly	O
increase	O
the	O
effectiveness	O
of	O
speech-recognition	B-Application
software	O
.	O
</s>
<s>
The	O
sentence	O
"	O
Five	O
women	O
left	O
"	O
,	O
which	O
could	O
be	O
phonetically	O
transcribed	O
as	O
[faɪvwɪmɘnlɛft],	O
is	O
marked	O
since	O
neither	O
/vw/	O
in	O
/faɪvwɪmɘn/	O
or	O
/nl/	O
in	O
/wɪmɘnlɛft/	O
are	O
allowed	O
as	O
syllable	B-General_Concept
onsets	O
or	O
codas	O
in	O
English	O
phonotactics	O
.	O
</s>
<s>
Infants	O
are	O
one	O
major	O
focus	O
of	O
research	O
in	O
speech	B-General_Concept
segmentation	I-General_Concept
.	O
</s>
<s>
Though	O
much	O
more	O
research	O
needs	O
to	O
be	O
done	O
on	O
the	O
exact	O
processes	O
that	O
infants	O
use	O
to	O
begin	O
speech	B-General_Concept
segmentation	I-General_Concept
,	O
current	O
and	O
past	O
studies	O
suggest	O
that	O
English-native	O
infants	O
approach	O
stressed	O
syllables	B-General_Concept
as	O
the	O
beginning	O
of	O
words	O
.	O
</s>
<s>
It	O
seems	O
that	O
infants	O
also	O
show	O
some	O
complexity	O
in	O
tracking	O
frequency	O
and	O
probability	O
of	O
words	O
,	O
for	O
instance	O
,	O
recognizing	O
that	O
although	O
the	O
syllables	B-General_Concept
"	O
the	O
"	O
and	O
"	O
dog	O
"	O
occur	O
together	O
frequently	O
,	O
"	O
the	O
"	O
also	O
commonly	O
occurs	O
with	O
other	O
syllables	B-General_Concept
,	O
which	O
may	O
lead	O
to	O
the	O
analysis	O
that	O
"	O
dog	O
"	O
is	O
an	O
individual	O
word	O
or	O
concept	O
instead	O
of	O
the	O
interpretation	O
"	O
thedog	O
"	O
.	O
</s>
<s>
Language	O
learners	O
are	O
another	O
set	O
of	O
individuals	O
being	O
researched	O
within	O
speech	B-General_Concept
segmentation	I-General_Concept
.	O
</s>
