| 吴宋体,陈家虎.酒店在线评论在TextCNN 和SVM 方法下的情感分类研究[J].情报工程,2026,(1):059-070 |
| 酒店在线评论在TextCNN 和SVM 方法下的情感分类研究 |
| Research on Sentiment Classification of Hotel Online Reviews Using TextCNN and SVM Methods |
| |
| DOI： |
| 中文关键词: 情感分析；TextCNN；在线评论；SVM；酒店服务 |
| 英文关键词: Sentiment Analysis; TextCNN; Online Reviews; SVM; Hotel Services |
| 基金项目: |
| 作者 | 单位 | | 吴宋体 | 桂林学院工商管理学院 桂林 541004 | | 陈家虎 | 广州中广计算机科技有限公司 广州 510000 |
|
| 摘要点击次数: 4 |
| 全文下载次数: 2 |
| 中文摘要: |
| [目的/意义]在线评论中蕴含丰富的用户情感信息，对酒店行业服务优化与管理决策具有重要意义。本文探讨了TextCNN 与SVM 两种情感分类方法在中文酒店评论中的应用效果。[方法/过程]基于携程平台的真实酒店评论数据，经过数据预处理与词向量化处理，分别构建并训练了TextCNN 和SVM 情感分类模型，通过准确率、精确率、召回率和F1 值等指标评估模型性能，并分析了错误分类样本。[局限]训练数据量较小，可能影响模型的泛化能力。[结果/结论]实验结果表明，TextCNN 和SVM 在测试数据中分别达到了96% 和93%的准确率，其中TextCNN在验证集中的准确率（86%）优于SVM（83%）。错误样本分析表明，两种模型在处理复杂或中性表达情感时存在误判，TextCNN对复杂语境的理解优于SVM。总体而言，研究验证了深度学习与传统机器学习在酒店评论情感分类中的有效性与差异，为酒店行业情感分析模型选择提供了实证依据。 |
| 英文摘要: |
| [Objective/Significance] Online reviews contain rich user sentiment information, which is of great significance for service optimization and management decision-making in the hotel industry. This paper explores the application effects of two sentiment classification methods, TextCNN and SVM, in Chinese hotel reviews. [Methods/Processes] Based on real hotel review data from the Ctrip platform, TextCNN and SVM sentiment classification models were constructed and trained after data preprocessing and word vectorization. Model performance was evaluated through metrics such as accuracy, precision, recall, and F1-score, and misclassified samples were analyzed. [Limitation] The small volume of training data may affect the generalization
capability of the models. [Results/Conclusions] Experimental results demonstrate that TextCNN and SVM achieved accuracies of 96% and 93% on the test data, respectively, with TextCNN outperforming SVM on the validation set (86% compared to 83%). Misclassified sample analysis indicates that both models exhibit misjudgments when processing complex or neutral expressions
of sentiment, and TextCNN demonstrates superior understanding of complex contexts compared to SVM. Overall, the study validates the effectiveness and differences between deep learning and traditional machine learning in the sentiment classification of hotel reviews, providing an empirical basis for model selection in hotel industry sentiment analysis. |
|
查看全文
查看/发表评论 下载PDF阅读器 |
| 关闭 |
|
|
|
