现代信息科技2024，Vol.8Issue(7)：141-145,150,6.DOI:10.19850/j.cnki.2096-4706.2024.07.029
基于预训练语言模型的旅游评论文本方面级情感分析研究
Aspect-based Sentiment Analysis Research of Tourism Review Text Based on Pre-trained Language Models
摘要
Abstract
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.关键词
语言模型/提示学习/方面级情感分析/预训练模型Key words
language model/prompt learning/aspect-based sentiment analysis/pre-trained model分类
信息技术与安全科学引用本文复制引用
谢宇欣,肖克晶,曹少中,张寒,姜丹..基于预训练语言模型的旅游评论文本方面级情感分析研究[J].现代信息科技,2024,8(7):141-145,150,6.基金项目
基于深度学习的虚假新闻检测关键技术研究(27170123034) （27170123034）
