METHOD FOR CLASSIFYING EMOTIONS OF SPEECH IN CONVERSATION BY USING SEMI-SUPERVISED LEARNING-BASED WORD-BY-WORD EMOTION EMBEDDING AND LONG SHORT-TERM MEMORY MODEL

摘要

A method for classifying emotions of a speech in a conversation by using semi-supervised learning-based word-by-word emotion embedding and a long short-term memory (LSTM) model, according to an embodiment of the present invention, comprises: a word-by-word emotion embedding step for tagging an emotion to each word in an input speech of conversation data by referring to a word emotion dictionary in which a corresponding basic emotion is tagged to each word for learning; a step for extracting an emotion value of the input speech; and a step for classifying emotions of the speech in consideration of changes in emotions in a conversation occurring in a messenger client, on the basis of the LSTM model, by using the extracted emotion value of the speech as an input value for the LSTM model. According to the present invention, appropriate emotions can be classified by recognizing changes in emotions in a conversation using natural language.