Research On The Method Of Mining Movie Critic Subjects Based On Sentiment Analysis

With the development of the Internet age,people can comment on movies through Internet platforms.By analyzing movie reviews,researchers can obtain meaningful latent emotional or topical information in the comments.Due to its self-learning ability and its advantages in large-scale data volume,deep learning methods have achieved good results in sentiment analysis by methods such as recurrent neural networks and convolutional neural networks.But these are all supervised learning models,and the results of the model training process are inseparable from the amount and quality of labeled text data,so mining more meaningful information from a certain amount of labeled text requires better solutions.At the same time,most of the researches that mine the information in the comments basically only focus on one field,that is,focus on sentiment analysis or topic mining.Few studies integrate the two into analysis,and a small number of studies also firstly perform topic mining and then perform sentiment analysis on the extracted topics.So it has certain application value to separate the mixed comments of different emotional tendencies.In the process of sentiment analysis,the key features in the text are not evenly distributed throughout the text,and it is easy to reduce the accuracy of text classification due to the loss of key features.This paper proposes a Bi GRU-CNN dual-channel sentiment analysis model based on attention mechanism.First,the input text is represented by word embeddings,and then Bi GRU is used to extract the time series features in the movie review data.At the same time,a self-attention mechanism is introduced into the hidden layer of the model.Extract local features between words in reviews using an improved two-layer CNN structure.Then,when the two extracted features are spliced and fused,an attention mechanism is introduced to assign global weights.By comparing the similarity of the feature vectors extracted by the two models and assigning different attention,the relationship between the two features is obtained,and the features are better fused.Through comparative experiments show that the sentiment analysis model proposed in this paper has better effect than other models in the evaluation indicators,and then verifies the validity and application value of the model in this paper.Aiming at the problem that movie review data with different emotional tendencies are chaotically mixed together,the Bi GRU-CNN dual-channel sentiment analysis model based on the attention mechanism proposed in this paper is used to perform sentiment analysis on the movie review data.According to the sentiment score obtained by the analysis,the review data is divided into different sentiment polarity intervals.Then,the topic mining of movie reviews is realized through the LDA topic model,and the optimal topic number is obtained by using the perplexity method.Finally,topic modeling is performed to extract topic words,and combined with movie review analysis to determine the meaning of topic words.By comparing and analyzing the subject words in different emotional polarity intervals,the potential topic information and hot topics of movies in different emotional polarity intervals are finally obtained.