Computer Engineering and Applications ›› 2023, Vol. 59 ›› Issue (2): 135-142.DOI: 10.3778/j.issn.1002-8331.2107-0144

• Pattern Recognition and Artificial Intelligence • Previous Articles     Next Articles

Personalized Product Review Summary Generation Based on TRF-IM Model

GAO Weijun, ZHU Jing, ZHAO Huayang, LI Lei   

  1. School of Computer and Communication, Lanzhou University of Technology, Lanzhou 730050, China
  • Online:2023-01-15 Published:2023-01-15

基于TRF-IM模型的个性化酒店评论摘要生成

高玮军,朱婧,赵华洋,李磊   

  1. 兰州理工大学 计算机与通信学院,兰州 730050

Abstract: Aiming at the problems of insufficient contextual understanding of reviews, insufficient parallelism and long-distance text dependence in the process of generating a summary of the traditional model of hotel review summary generation, a personalized hotel review summary generation method based on TRF-IM(improved mask for transformer) model is proposed. This method uses the Transformer decoder structure to model the review summary task. By improving the masking method in the structure, the source review content can better learn the contextual semantic information. At the same time, user-type personalized word feature information is introduced to generate high-quality personalized hotel review summaries that meet user needs. The experimental results show that the proposed model achieves higher scores on the ROUGE indicator than the traditional model, and generates high-quality personalized hotel review summaries.

Key words: personalized hotel reviews, generative summary, Transformer, masking method, language mode

摘要: 针对传统的酒店评论摘要生成模型在生成摘要过程中存在对评论的上下文理解不够充分、并行能力不足和长距离文本依赖缺陷的问题,提出了一种基于TRF-IM(improved mask for transformer)模型的个性化酒店评论摘要生成方法。该方法利用Transformer译码器结构对评论摘要任务进行建模,通过改进其结构中的掩码方式,使得源评论内容都能够更好地学习到上下文语义信息;同时引入了用户类型的个性化词特征信息,使其生成高质量且满足用户需求的个性化酒店评论摘要。实验结果表明,该模型相比传统模型在ROUGE指标上取得了更高的分数,生成了高质量的个性化酒店评论摘要。

关键词: 个性化酒店评论, 生成式摘要, Transformer, 掩码方式, 语言模型