Application And Research Of Convolution Neural Network In Emotion Classification

Emotional classification aims to discover the emotional polarity of viewers and has been an important task in natural language processing.In order to solve the problem of short text,the traditional methods are too dependent on the emotional dictionary and manually set the characteristic engineering,etc.The paper proposes emotion classification model which is based on the two-norm linear support vector machine,the deep convolution neural network and parallel convolution neural network.The two models use the depth network structure and the parallel network structure to extract the emotion local characteristics respectively.Using the two-norm linear support vector machine optimization function which is different from the classical convolution neural network model,The gradient problem of parameter optimization process is solved.by using the Skip-gram pre-training vector model in word2 vec to express the sparse expression of the short text and the emotional feature is compressed by the dynamic pooling function.In the real network review data set and the classical method Quantitative comparison,the accuracy and recall rate of the two classification models is greatly improved,and the validity of the two-norm support vector machine parameter optimization is verified by the change of the weight update,and the best model performance is obtained by adjusting the penalty coefficient.