<s>
Emotion	B-Application
recognition	I-Application
in	I-Application
conversation	I-Application
(	O
ERC	O
)	O
is	O
a	O
sub-field	O
of	O
emotion	O
recognition	O
,	O
that	O
focuses	O
on	O
mining	O
human	O
emotions	O
from	O
conversations	O
or	O
dialogues	O
having	O
two	O
or	O
more	O
interlocutors	O
.	O
</s>
<s>
The	O
datasets	O
in	O
this	O
field	O
are	O
usually	O
derived	O
from	O
social	O
platforms	O
that	O
allow	O
free	O
and	O
plenty	O
of	O
samples	O
,	O
often	O
containing	O
multimodal	B-General_Concept
data	O
(	O
i.e.	O
,	O
some	O
combination	O
of	O
textual	O
,	O
visual	O
,	O
and	O
acoustic	O
data	O
)	O
.	O
</s>
<s>
Approaches	O
to	O
ERC	O
consist	O
of	O
unsupervised	B-General_Concept
,	O
semi-unsupervised	O
,	O
and	O
supervised	B-General_Concept
methods	O
.	O
</s>
<s>
Popular	O
supervised	B-General_Concept
methods	O
include	O
using	O
or	O
combining	O
pre-defined	O
features	O
,	O
recurrent	B-Algorithm
neural	I-Algorithm
networks	I-Algorithm
(	O
DialogueRNN	O
)	O
,	O
graph	O
convolutional	O
networks	O
(	O
DialogueGCN	O
)	O
,	O
and	O
attention	O
gated	O
hierarchical	O
memory	O
network	O
.	O
</s>
