Research on Emotion Tendency Analysis Model Based on Deep Learning

With the rapid development of the Internet and the rise of major rating websites, consumers will share their experiences of consumption and experience through microblogs, comment websites and other channels, especially in the food, shopping, film and television industries. The highly praised comments generated after the movie premiere can directly affect the purchase decisions of subsequent potential consumers. On the one hand, positive premiere reviews are a secondary promotion for the movie, which can attract more consumers to watch. On the other hand, excessive negative reviews not only cause losses to the single box office, but also lower the producer's production level and actors' acting skills in the minds of consumers, damaging the team's reputation and causing long-term consumer loss. Therefore, conducting sentiment analysis on real-time online comment data can not only help companies in need to respond promptly to negative public opinion in real time and develop corresponding countermeasures; By increasing the granularity of consumer review data analysis, deep mining and analysis can be conducted to discover real-time consumption trends and preferences of consumers, which can be applied to multiple business fields such as user interest mining and personalized product recommendations. This article is based on movie review data on Douban website, using sentiment dictionaries LSTM Two different models were compared, and text collection and preprocessing techniques such as data cleaning, Chinese word segmentation, and stop word filtering were adopted. Based on this technology, a model for online film reviews on Douban was established through an sentiment dictionary, and the accuracy of the basic model was verified using actual data; Subsequently, the LSTM algorithm will be used to implement the sentiment classification model. The model will also be validated using actual data to calculate its accuracy. The advantages and disadvantages of the two models will be compared and analyzed, and a summary will be made.