The Application of An Optimized Convolutional Neural Network Model in Film Criticism

Abstract

Constructing a model of online film and television commentary sentiment classification can effectively guide film and television producers to comprehensively understand the audience acceptance of film and television works, and improve it. Traditional methods based on sentiment lexicon and machine learning exist in a series of Insufficient: ignore context semantics, too single word, sparse features, etc. Based on the existing convolutional neural network model, this paper systematically optimizes its internal structure, and proposes a NCNM (New Convolutional Neural Network model) model based on multi-sliding window and new pooling method, and uses feature vectors to cluster feature words. . In this paper, the Stanford SST dataset and Cornell MRD dataset are used to verify the classification effect of the proposed model. The experimental results show that ncnnm has a certain improvement in the accuracy of the emotional classification of short text video reviews compared with the existing mainstream methods..