Research On Analysis Method Of Emotional Cues In Conversation

With the development of science and technology,modern society has been a society in which artificial intelligence technology is closely combined with daily life.People’s life has become simple and convenient because of artificial intelligence technology.In the field of artificial intelligence,the research on analysis method of emotional cues in conversation has attracted more and more attention from scholars,it makes human-computer interaction system more humanized.The research on analysis method of emotional cues in conversation is mainly based on the emotion recognition in the conversation and the anomaly detection of time series.Among them,the emotion recognition in the conversation plays an extremely important role.This paper designs two emotion recognition models in conversation to extracts emotional clues from the aspects of integrity,continuity and variability,and puts forward an analysis method of emotional clues in the conversation,which is studied from the aspects of integrity,continuity and variability.In general,the main research work of this paper is as follows:1)In the sentence structure,a word or phrase often plays the role of emotional transition or emotional progression,but the order of phrases in time coordinates is often not continuous.Previous models mostly used convolutional neural networks or gated recurrent units to extract the emotional features of sentences according to time sequence,and did not pay close attention to the logical relationship between multiple parts of sentences segmented by phrases.These methods ignore the subtle emotional cues in the sentence structure,such as emotional transformation and emotional progression,which usually represent strong emotional changes,which is conducive to the study of emotional cues in the variability level of the conversation.In order to solve the problem,this paper proposes an emotion recognition model TANTP in the conversation.TANTP integrates the context information and the structure of the binary constituency analysis tree into the model structure.Binary constituency analysis tree can analyze the sentence structure and emotional logic,and capture the correct emotional features including complex emotional transformation and emotional progression for the context memory bank.The experimental results on two public datasets show that the TANTP model has better performance than other advanced models.2)The attention mechanism based on the binary constituency analysis tree is conducive to the study of emotional cues at the variability level.In order to deeply explore the emotional clues in the conversation from the aspects of integrity and continuity,this paper starts with changing the conversation context information fusion mechanism,and puts forward the emotion recognition model MPMTP in the conversation.The context fusion mechanism of emotion recognition model in the conversation depends on the identity information and location information of conversation participants to a certain extent.Therefore,MPMTP model designs four different self-attention mechanisms according to the identity information and location distance information of speaker and listener in the conversation.These four types of self-attention mechanisms can more carefully grasp the relevant links between conversation participants and build a new context information fusion framework,which is conducive to the research on the integrity and continuity of emotional clues in the conversation.This paper will show the performance of MPMTP model on the two public datasets.3)In order to improve the rationality of the experimental results of emotional cue analysis method,this paper labels the conversations in the first 20 episodes of the first season of the TV series Friends,which greatly expands the corpus of this paper.In order to make the research results of emotional clues more scientific and intuitive,this paper uses the emotional recognition model to generate sentence vectors with rich emotional features,and inputs these sentence vectors into the anomaly detection model of time series to mine the clues of emotional in the the conversation through the anomaly detection model of time series.The whole experimental process is a new attempt,which puts forward the research method of the emotional cues in the conversation for the first time.According to the test results of corpus,this paper studies the analysis method of conversation emotional cues from three aspects:integrity,continuity and variability.