基于在线评论的概率语言多属性评价方法

Probabilistic Language Multi-Attribute Evaluation Method Based on Online Reviews

随着互联网技术的发展和普及,语言决策涉及的领域愈发广泛,评价某一事物的方式越来越多样化,采用调查问卷或者专家决策的方式来进行评价已经无法满足某些业务需求。在线评论文本中包含了大量关于产品或服务的用户体验信息,但是数据量大且语言结构复杂增加了决策者给出评价结果的难度。因此,本文基于评论文本建立了一种概率语言多属性决策(PL-MCDM)模型来评估产品或服务。首先从文本中提取主题词构建属性词词典,通过关键词提取和情感分析模型得到评论中的关键词词对和情感倾向;其次,考虑到情感分析对评论者意见表达的置信度,采用概率语言术语集来扩展情感分析的结果,最终对产品或服务进行综合排名。最后,本文将所提出的模型应用到一个餐厅质量评估的案例中,结合现有结果证明了本文方法的可行性和有效性。

With the development and popularization of Internet technology, the fields of language decision-making are more and more extensive, and the ways of evaluating something are more and more diversified. The traditional expert decision-making can not meet some daily needs. Online reviews contain a lot of user experience information, but the amount of data is large and the language structure is complex, which increases the difficulty of decision-makers to give the evaluation results. Therefore, this paper establishes a probabilistic linguistic multiple attribute decision making (pl-madm) model based on comment text to evaluate products or services. Firstly, the subject words are extracted from the text to construct the attribute word dictionary, and the keyword pairs and emotional tendencies in the comments are obtained by keyword extraction and sentiment analysis model;secondly, considering the confidence degree of sentiment analysis on the opinion expression of the reviewers, probabilistic language term set is used to expand the result of sentiment analysis, and finally the products or services are ranked comprehensively. Finally, the proposed model is applied to a restaurant quality assessment case, and the results show that the method is feasible and effective.