Aspect-based Sentiment Analysis Research of Tourism Review Text Based on Pre-trained Language Models
In order to promote consumption in the tourism industry and economic development,we analyze the scenic spot comment texts published by tourists on online platforms,and deeply explore the fine-grained emotional information in them,in order to better cater to the preferences of tourists.In actual scenarios,a sentence may involve multiple entity words,making it difficult to accurately identify their corresponding emotional attribute relationships.Moreover,there are issues of scarcity and imbalanced samples in the dataset of tourism scenarios.A pre-trained language model based on Deep Learning and prompt knowledge is constructed.Two sub tasks are jointly trained by constructing a discrete prompt template,and data augmentation is performed on a few samples in the dataset.At the same time,different weights are set for the loss function during the training phase.The experimental results show that the model has achieved significant results on the tourism review text dataset and the public dataset SemEval2014-Restarantt,with F1 values reaching 80.81%and 83.71%,respectively,which helps tourism institutions to achieve personalized analysis of each city's scenic spots.
language modelprompt learningaspect-based sentiment analysispre-trained model
国家科技期刊平台
NSTL
万方数据
维普
