Abstract
Using technology for sentiment analysis in the travel industry can extract valuable insights from customer reviews. It can assist businesses in gaining a deeper understanding of their consumers' emotional tendencies and enhance their services' caliber. However, travel-related online reviews are rife with colloquialisms, sparse feature dimensions, metaphors, and sarcasm. As a result, traditional semantic representations of word vectors are inaccurate, and single neural network models do not take into account multiple associative features. To address the above issues, we introduce a dual-channel algorithm that integrates convolutional neural networks (CNN) and bi-directional long and short-term memory (BiLSTM) with an attention mechanism (DC-CBLA). First, the model utilizes the pre-trained BERT, a transformer-based model, to extract a dynamic vector representation for each word that corresponds to the current contextual representation. This process enhances the accuracy of the vector semantic representation. Then, BiLSTM is used to capture the global contextual sequence features of the travel text, while CNN is used to capture the richer local semantic information. A hybrid feature network combining CNN and BiLSTM can improve the model's representation ability. Additionally, the BiLSTM output is feature-weighted using the attention mechanism to enhance the learning of its fundamental features and lessen the influence of noise features on the outcomes. Finally, the Softmax function is used to classify the dual-channel fused features. We conducted an experimental evaluation of two data sets: tourist attractions and tourist hotels. The accuracy of the DC-CBLA model is 95.23% and 89.46%, and that of the F1-score is 97.05% and 93.86%, respectively. The experimental results demonstrate that our proposed DC-CBLA model outperforms other baseline models.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have