Research On Text Sentiment Analysis Method Based On Deep Learning Technology

In the past ten years,thanks to the rapid development of the economy,China's Internet has been producing massive amounts of text data.How to use natural language processing to extract useful information from these texts has become an important research topic.Sentiment analysis is an important branch of natural language processing,which aims to use techniques such as semantic mining to analyze the impact of sentiment classification on commodities,human events,and problem processing.At present,there are three common sentiment analysis methods in the industry: dictionary-based methods,machine learning-based methods,and deep learning-based methods.As the complexity of online text data continues to increase,many online texts contain more than one application field.In order to allow the sentiment analysis model to be applied in more refined fields,this paper proposes a more comprehensive,and can also maintain cross-domain application.A sentiment analysis method with high classification accuracy,this method combines local and global features,and then performs semantic enhancement based on the importance of keywords,and finally conducts experiments on three different datasets of social,online shopping and dining.The comparison results are analyzed to verify the generality and effectiveness of the algorithm proposed in this paper.In the above sentiment analysis research process,we often do not make sufficient use of the dataset,and most models only use two labels,comment text and sentiment polarity,resulting in a waste of data resources.Moreover,in actual production,it is also a common method for researchers to improve and optimize the model in a specific field to improve the accuracy.Therefore,this paper continues to optimize the model in the previous section.While integrating structured data,a soft attention mechanism is also introduced.Balance the weights of multivariate features.This article uses crawler technology to pull multiple types of song comment data from the Net Ease cloud platform.The song reviews,song categories and other information obtained after preprocessing are used as self-built data sets for ablation experiments and comparison experiments.The results show that the classification accuracy of the sentiment analysis model proposed in this paper can be further improved by integrating the external information of song type.Based on the work of this paper,it can be seen that deep learning technology is feasible and effective in Chinese text sentiment analysis.Exploring the application of text sentiment analysis in different fields and the coordinated development between deep learning and multidisciplinary are important research directions in the future.