Research on Text Sentiment Analysis based on Word2Vec and Improved DPCNN Model

With the rise of social media and online reviews, sentiment analysis has become a key focus in natural language processing, widely used for analyzing user feedback and monitoring public opinion. Conventional sentiment analysis approaches often depend on handcrafted features or simple deep learning models, which struggle to effectively grasp deep semantic relationships. In analyzing long texts, capturing global context and handling long-range dependencies remains a significant challenge. By leveraging Word2Vec, the model transforms text into vector representations to strengthen feature representation. Furthermore, incorporating self-attention into DPCNN enhances its ability to capture long-range dependencies and global semantics. Using IMDB dataset as the dataset of the model in this study, based to the research, the result accuracy percentage terms is 90.17%. which indicates that this model is improved over the traditional model. Findings indicate that the proposed model, which combines Word2Vec with an improved DPCNN, enhances sentiment classification performance and addresses the shortcomings of conventional approaches.