Research On Sentiment Analysis Of Social Platform Reviews Based On Deeplearning

In recent years,more and more people have expressed their opinions on social platforms.Every day,social platforms such as Sina Weibo and other social platforms will generate a lot of people's emotional expressions of hot events,and these information often exist in the form of text.Obtaining information about the emotional orientation in comments on social platforms and understanding people's attitudes to events have very important application value in rumor control,marketing,and public opinion monitoring.The traditional sentiment analysis method has great limitations,it is difficult to identify the more obscure sentiment information in the text,and it cannot adapt to the ever-changing information in today's era.With the rise of deep learning,using deep learning to improve the effect of text sentiment analysis has become a research hotspot.Therefore,this paper designed two deep learning models to conduct sentiment analysis research on social platform reviews.This article first uses the Scrapy framework to crawl comment information from the Weibo social platform as a model data set,and then preprocesses these comment data,including Chinese word segmentation,part-of-speech tagging,and stop word removal operations.An improved Word2 vec model is proposed to perform word vector processing on these data,that is,to convert text sentences into corresponding word vector matrices and input them into the deep learning classification model.Finally,a comparative experiment is carried out on the improved Word2 vec model to verify the advantages of the improved Word2 vec model.Text convolutional neural network Text CNN can obtain the local feature representation of sentences.The BiGRU model of bidirectional gated circulation unit can obtain the time series relationship between text words and sentences,and extract the global features of the text.This paper designs the Text CNN-BiGRU model for sentiment analysis.The model combines the advantages of the two networks to improve the accuracy of text sentiment analysis.In order to optimize the analysis effect of the model,multiple sets of comparative experiments are also carried out on the determination of parameters,including parameters such as convolution kernel size,batch_size value,and Dropout value.In order to verify the effectiveness of the model,the Text CNN-BiGRU model and five sets of deep learning network models were compared in multiple dimensions.The experimental results prove that the Text CNN-BiGRU model performs well on text sentiment analysis tasks.In addition,in order to make the classification model more perfect,this paper also introduces a self-attention mechanism based on the Text CNN-BiGRU model to increase the weight of important text features and build a new model SAC-BiGRU.In order to verify the effectiveness of the SAC-BiGRU model with self-attention mechanism,the model was compared with four groups of deep learning network models in multiple dimensions.The results prove that the SAC-BiGRU model has good text sentiment analysis capabilities.