Dialogue Emotional Recognition Based on Comparative Topic Model and Label Guidance
In a dialogue system,dialogue emotion recognition aims to predict the emotional labels for each statement in a conversation.This task is influenced by various factors such as the topic of the dialogue,emotional label information and so on.To address this issue,a new framework called Topic-aware Label Graph Network(TLGN)has been proposed.This model uses topic model to extract semantic topic distributions related to emotions and learns label-guided text representations firstly.Then,it fuses two representations of different granularity,used as semantic input to the text and input into the classification model for emotional prediction of dialogue statements.Finally,experimental results on four publicly available datasets show that the model method outperforms the benchmark methods.
国家科技期刊平台
NSTL
万方数据
