Research On Extraction Method Of Emotion And Cause Clause

Text sentiment analysis(also known as opinion mining)has always been a research hotspot in the field of natural language processing(NLP).Therefore,the current research focus is from the increasingly mature text emotion analysis to the mining of the causes of text emotion,that is,from "knowing what it is" to "knowing why it is",that is,to extract the causes of text emotion.The traditional task of emotion cause extraction is to extract the corresponding cause based on the condition of known emotion expression.Today,with the rise of deep learning,when the traditional cause extraction task is combined with deep learning,emotional information needs to be labeled in the dataset,which increases the labor cost and limits its application in the real world.Therefore,a new task of emotion cause extraction appears:extracting emotion information and causes at the same time.The main research content of this paper is to combine deep learning technology to predict the text emotion clause without given emotion tag,and extract the cause clause at the same time.Firstly,the emotion prediction and cause extraction model(EPCEM)is proposed.In this model,the local semantic information,context semantic information and the key part of emotion expression are fused to obtain better clause features for emotion clause prediction.And the cause clause is extracted by combining the prediction result of emotion clause with the feature of clause.The experimental results show that the model achieves good results in both emotion clause prediction and cause clause extraction tasks,and the cause extraction effect is still better than most traditional models in the case of no given text emotion label.In order to further improve the extraction effect of emotion clause and cause clause,an interactive emotion clause and reason clause extraction model is proposed based on the interaction between emotion and reason.The model adds the interactive attention matrix to extract the relevance between emotion and cause,so as to improve the extraction effect of emotion clause and cause clause.The experimental results show that the interaction attention matrix can significantly improve the effect of the model.