Mining Attribute-Specific Ratings from Reviews of Cosmetic Products
Mining Attribute-Specific Ratings from Reviews of Cosmetic Products
复制标题
从化妆品评论中挖掘特定属性的评级
DOI：
10.1007/978-981-10-3950-8_8
复制
发表时间：
2017
期刊：
影响因子：
--
通讯作者：
Byungkyu Kang
中科院分区：
文献类型：
--
作者：
Yuuki Matsunami;Mayumi Ueda;Shinsuke Nakajima;Takeru Hashikami;John O'Donovan;Byungkyu Kang
在化妆品领域，许多在线卖家支持用户提供的产品评论。事实证明，评论对产品转化率有着深远的影响。化妆品的评论在购买决定中具有特别重要的意义，因为它们的个人性质，特别是因为不合适的产品可能会引起刺激。在本文中，我们提出了一种方法自动评分的化妆品项目审查文本的各个方面的基础上策划的字典的表达从语料库的真实的世界在线评论。结果和讨论的用户实验，以评估该方法。特别是，我们发现共现方法提高了评论的覆盖率，并且我们的自动方法预测了手动注释的地面真实属性，准确率为79%。
In the cosmetics domain, many online sellers support user-provided product reviews. It has been shown that reviews have a profound effect on product conversion rates. Reviews of cosmetic products carry particular importance in purchasing decisions because of their personal nature, and particularly because of the potential for irritation with unsuitable products. In this paper, we propose a method for automatic scoring of various aspects of cosmetic item review texts based on a curated dictionary of expressions from a corpus of real world online reviews. Results and discussion of a user experiment to evaluate the approach are presented. In particular, we find that a co-occurrence approach improved coverage of reviews, and that our automated approach predicted attributes in manually annotated ground truth with an accuracy of 79%.