Periodic insight: Multilingual reputation generation system through daily opinion mining analysis

Abstract

The global trend of individuals expressing their opinions on Twitter has led to a substantial number of user-generated reviews across various brands, products, and services. As a result, there is a growing need for automated systems capable of analyzing and interpreting this extensive content. In response, reputation generation systems have been developed to extract valuable insights from both textual and numerical reviews. However, many of these systems have significant limitations. Firstly, most of them are limited to processing English text, which poses a barrier for analyzing reviews in other languages. Also, they are incapable of handling immediate data influx, they often fall short in providing up-to-date and accurate reputation assessments. Therefore, we propose a two-phase system for generating accurate reputation values. In the first phase, data preparation, the system incorporates review translation to English, spam filtering, and sarcasm detection to address limitations of language processing and enhance data quality. This prepares the data for the second phase, reputation generation, which utilizes state-of-the-art, aspect-based sentiment analysis techniques, offering a novel approach to calculating reputation by considering specific aspects of products or services. Experimental results conducted on multiple datasets show the efficacy of the proposed system compared with previous reputation generation systems.