A Deep Learning Model for Analyzing Smartwatch Reviews Based on BO-BiLSTM

In recent years, with the advancements in deep learning models, the influence of artificial intelligence technology in social life has been increasingly prominent. Smartwatches, as embedded AI devices, are becoming a key gateway to the era of big data. However, the enrichment of smartwatch functionalities and their widespread usage have generated a vast amount of user feedback data. This paper aims to leverage big data technology to mine and analyze online user reviews of Huawei-branded smartwatches, utilizing deep learning-based natural language processing techniques for feature selection and sentiment value prediction to understand user needs and guide product optimization and innovation. Using the JD.com platform as the data source, we crawled online review texts and conducted initial preprocessing such as text deduplication and mechanical word compression. We then performed topic analysis and feature extraction on the text and calculated sentiment scores for each feature. Finally, we employed the BO-BiLSTM model to explore the sentiment value fluctuation trends of various attribute features. Taking Huawei smartwatches as a case study, the results indicate that the positive sentiment values for "hardware features" are on an upward trend, while the sentiment values for "user experience features" are on a downward trend. These findings provide valuable insights for enterprises to better understand market changes and technological iterations for enhanced development.