中国烟草学报2025，Vol.31Issue(5)：146-154,9.DOI:10.16472/j.chinatobacco.2024.T0387
基于微调大语言模型的互联网卷烟消费评价文本分类研究
Research on internet cigarette consumption evaluation text classification based on fine-tuning large language model
骆宇峰 1黄斌1
作者信息
- 1. 九江市烟草公司柴桑分公司,江西省九江市柴桑区江州大道 332000
- 折叠
摘要
Abstract
This paper aims to study a cigarette consumer evaluation text classification method based on fine-tuned large language models to achieve accurate classification of cigarette product reviews.Addressing the challenges faced by traditional deep learning models in processing internet Chinese text,this paper designs an artificial intelligence model that can simultaneously consider sentiment classification,topic classification,and natural language inference.The research method includes constructing a dataset containing five types of labels and using supervised fine-tuning and reinforcement fine-tuning to train the large language model.The experimental results show that the Micro F1 score of this model on the test set is improved by up to 26.72%compared to the baseline model,effectively improving the classification accuracy and flexibility of cigarette product reviews.The application of this model can help enterprises quickly identify their shortcomings from internet cigarette consumer reviews,improve product improvement efficiency,optimize marketing strategies,and maintain an advantage in the fiercely competitive market.关键词
烟草/文本分类/情感分析/主题分类/大语言模型/深度学习Key words
tobacco/text classification/sentiment analysis/topic classification/natural language inference/large language mode/deep learning引用本文复制引用
骆宇峰,黄斌..基于微调大语言模型的互联网卷烟消费评价文本分类研究[J].中国烟草学报,2025,31(5):146-154,9.
