Bullet Screen Short Text Sentiment Analysis Algorithm

Abstract

Bullet screen can express the feeling of the audience when watching video. If emotional analysis and research are carried out on it, it is helpful to improve the accuracy of user recommendation system. The existing emotional analysis method of bullet screen text separates emotional symbols from text information and ignores emotional expression of emotional symbols in bullet screen. To this end, an emotional symbol space multi attention convolutional neural network model (ES-MACNN) is proposed for video barrage sentiment analysis, the model selects emoji and kaomoji to construct emotional symbol space, uses attention mechanism to describe the important degree of emotional information, and obtains deeper emotional feature information to achieve the purpose of improving classification accuracy. Through experimental verification and analysis, the ES-MACNN model has significantly improved classification precision and recall rate compared with the traditional model.