“…Conditional Random Field To model the label consistency, i.e., the emotion transfer on a dialogue, a linear chain conditional random field is employed to yield final emotion tags of each utterance. Following the description of (Lample et al 2016), for an input set of utterances U = u 1 , u 2 , ..., u n and a sequence of tag predictions y = y 1 , y 2 , .., y n , y i ∈ 1, • • • , K (K is the number of emotion tags), the score of the sequence is defined as:…”