The Multi-label Text Classification Method for Jingdezhen Tourism Reviews Based on ERNIE-TextCNN Fusion with Label Information

Travelers can effectively plan their trips by referring to attraction reviews available on tourism websites. However, such reviews typically involve multiple dimensions, and tourism ticketing platforms often lack comprehensive keyword segmentation within reviews, resulting in inconsistent classification standards. To tackle this problem, we propose an ERNIE-TextCNN model that effectively integrates label information. The ERNIE model captures deep semantic features from tourism review texts, while TextCNN extracts local key information. Additionally, label information is innovatively incorporated into the model, effectively overcoming the limitations of traditional models, such as inadequate extraction of semantic features and insufficient utilization of label semantics. Experiments on the Jingdezhen tourism review dataset and CAIL2019 dataset confirm the proposed method's effectiveness. Experimental results indicate that incorporating label information significantly improves the accuracy of multi-label classification tasks, offering novel insights for multidimensional data mining of tourism reviews.