Sentiment analysis algorithm based on word embedding in text mining

Abstract

This study proposes a sentiment analysis algorithm model that combines word embedding and machine learning algorithms to address issues such as data shortage, low accuracy, and limited coverage. The proposed model uses the continuous bag-of-words word embedding method for word analysis and incorporates convolutional neural network, long and short-term memory neural network model, and attention mechanism to enhance the traditional word embedding method.The research results obtained using the English review dataset and the hotel review dataset, which contain an average of 10,000 entries, show that in the performance comparison of hotel review sentiment analysis tests, the highest value of the model loss function used in this study is 0.92. The accuracy, recall, and F1 value of the model are 90.56%, 91.65%, and 92.36%, respectively, which are 5.89%, 7.00%, and 6.74% higher than the lowest values in the convolutional neural network, long short term memory neural network, and convolutional long short term memory neural network. In the performance comparison for different emotional tendencies, the algorithm model used in this study achieved various indicator values of 94.68%, 93.86%, and 94.68% in sentiment analysis, respectively. The model with attention mechanism has a higher accuracy, with an accuracy of 96.4%. This shows that the algorithmic model used in the study is able to improve the accuracy and effectiveness of sentiment analysis to some extent. The research model is particularly suitable for sentiment analysis tasks in scenarios such as product reviews and hotel evaluations on e-commerce platforms, and can extract key emotional information from massive user reviews. This provides strong support for user sentiment analysis and data-driven decision-making in industries such as e-commerce and tourism.