SEOVER: Sentence-Level Emotion Orientation Vector Based Conversation Emotion Recognition Model

For the task of conversation emotion recognition, recent works focus on speaker relationship modeling but ignore role utterance's emotional tendency.In this paper, we propose a new expression paradigm sentence-level orientation vector to model potential correlation emotions between sentence vectors. Based it, design an recognition model, which extracts vectors from language and jointly learns dialogue sentiment analysis extracted identify speaker's during conversation. We conduct experiments two benchmark datasets compare them with five baseline models.The experimental results show that our has better performance all data sets.