基于UGC文本的动态多属性决策方法

Dynamic Multiple-Attribute Decision Making Method Based on UGC Text

摘要: 网络中的大量UGC文本包含了用户对产品的真实评价，如何利用评论文本对产品进行决策或排序越来越受到重视。本文提出了基于UGC的动态多属性决策方法。首先，面对非结构化的文本信息难以直接被机器理解，提出了属性信息提取方法。从文本中提取用户关注的产品属性，基于情感分析方法对属性评价值进行计算，并结合属性频次和情感分歧度指标定义了属性权重计算方法。然后，考虑到UGC文本内容的动态变化导致不同时段的信息对当前决策的参考价值不同，本文在多属性决策框架的基础上提出了区分时间窗口的动态决策模型，时间窗口的划分基于属性相似性，同时结合时间距离给出了时间权重计算方法。最后，运用本文提出的方法结合VIKOR排序算法对基于游记的旅游目的地排序问题进行了实验求解，说明了本文方法的可行性和实用性。

Abstract: A large number of UGC texts in the network contain users’ real evaluations of products. How to use the comment texts to make decisions or rank products has been paid more attention. This paper proposes a dynamic multi-attribute decision-making method based on UGC. First of all, in the face of unstructured text information that is difficult to be directly understood by the machine, an attribute information extraction method is proposed to extract the product attributes that users care about from the text, then calculate the attribute evaluation value based on the sentiment analysis method, and define the attribute weight combining the attribute frequency and sentiment divergence index. Secondly, considering that the dynamic changes of online review content will lead to different reference values of information in different periods for current decisions, this paper pro-poses a dynamic decision model to distinguish time windows based on the multi-attribute decision-making framework. The division of time windows is based on attribute similarity, and the calculation method of time weight is given in combination with time distance. Finally, the proposed method combined with the VIKOR sorting algorithm is used to solve the tourism destination sorting problem based on travel notes, which shows the feasibility and practicability of the proposed method.