A T1OWA fuzzy linguistic aggregation methodology for searching feature-based opinions

作者:

Highlights:

摘要

Online services such as Amazon, Tripadvisor, Ebay, etc., allow users to express sentiments about different products or services. Not only that, in some cases it is also possible to express sentiments about the different features characterizing those products or services. Most users express sentiments about individual features by using numerical values, which sometimes do not allow users to reflect properly what they are meaning and therefore they are misleading. To overcome this key issue and make users’ opinions in online services more comprehensive, a new methodology for representing sentiments using linguistic term sets instead of numerical values is presented. In addition, this methodology will allow to implement importance degrees on the different features characterizing users’ opinions. From both sentiments and importance of the features, the most important opinions for each user is derived via an aggregation step based on the Type-1 Ordered Weighted Averaging (T1OWA) operator, which is able to aggregate the corresponding fuzzy set representations of linguistic terms. Furthermore, the final output of the T1OWA based-search process can easily be interpreted by users because it is always of the same type (fuzzy) and defined in the same domain of the original fuzzy linguistic labels. A case study is presented where the T1OWA operator methodology is used to assess different opinions according to different user profiles.

论文关键词:Sentiment analysis,Feature opinion aggregation,Linguistic model,T1OWA aggregation operator

论文评审过程:Received 15 April 2019, Revised 12 October 2019, Accepted 14 October 2019, Available online 18 October 2019, Version of Record 16 January 2020.

论文官网地址:https://doi.org/10.1016/j.knosys.2019.105131