Sentiment Analysis Of Social Networks Based On Deep Learning

Deep learning(DL,Deep Learning)is an important field of emerging development in the field of computer science today.It is a further expansion on the basis of traditional machine learning(ML,Machine Learning),and thus is closer to higher-order artificial intelligence(AI,Artificial Intelligence)The goal.The basic principle of deep learning is to learn the inherent regular distribution of sample data and mining hidden sample features,so that in the learning process,the analysis and understanding of data such as images,text,and speech can be obtained.Finally,the deep learning model that is trained can be Have the ability to analyze and infer like people,solve classification or regression problems.Deep learning is a higher level on the basis of traditional machine learning.It is considered to be a more complex machine learning algorithm.Based on the sparse features of big data,it often achieves better results,such as file retrieval,data mining,and machine learning.Significant progress has been made in related fields such as translation,image classification,intelligent driving,and personalized recommendation.Deep learning learns to simulate human thinking through as many samples as possible,helps solve many complicated and tedious classification and prediction tasks,and greatly accelerates the development of artificial intelligence.Based on the direction of natural language processing,this article organizes and summarizes the latest research results based on deep learning at home and abroad,and briefly introduces the concepts and algorithms involved in deep learning from fully connected neural networks to classic convolutional neural networks and recurrent neural networks.,The latest recurrent neural network is optimized by adding the attention mechanism to solve the classification problem of social sentiment analysis,and the effect performance analysis is carried out at the theoretical level and the practical level through comparison.This article mainly focuses on the issue of sentiment analysis.The research work consists of the following two parts:(1)Transform on the basis of traditional CNN and LSTM network models,borrow the gated elements of LSTM from CNN,and input the vector of the gated convolution layer to the sentiment classifier layer for classification,and construct several layers with design A network with a predetermined number of neurons,as well as CNN and LSTM with special processing methods between layers complete the feature extraction work,and apply each model to the multi-classification problem of social network sentiment analysis,through these different convolutional neural networks and loops The neural network learns the characteristics and the recognition ability in the experiment to analyze the performance difference of each model.(2)By drawing on the idea of ??attention mechanism(Attention),a deep learning framework in which attention weights are added to gated convolution is constructed,and a special attention weight method is adopted to enhance semantic information;secondly,it integrates gated convolution and Attention also builds a model for aspect related semantics and complete semantic information of the text;finally,the vector of the gated convolutional layer is input to the sentiment classifier layer for classification,and the softmax function is used to complete multiple classifications,that is,negative,positive,and neutral,to calculate the emotion category the result of.(3)Through practice,apply the optimized convolutional neural network loop optimization network to single task and multi-task,compare other existing deep learning classifiers,and verify the optimized convolutional neural network and recurrent neural network in social network sentiment analysis The feasibility of the application.