Your Words Count: Investigating Inconsistency in Reviews from Multiple Online Sources Via Topic Modeling

摘要

Thanks to the development of Web 2.0, the shopping strategy of consumers has been changed significantly in recent years. Consumers are seeking others' shopping experiences and opinions on products from all kinds of platforms, such as e-commerce site, dedicated review sites, blogs, discussion groups, forums, and other social network services, rather than relying on their peers' word-of-mouth. However, being overwhelmed by the information on the internet nowadays, it is time-consuming, tedious and even infeasible for consumers to go through every single review from one online source, let alone the reviews about the same product from other sources. Thus consumers usually rely on reviews from a single source helping them make purchase decisions. However, an early research showed the overall opinion orientations of the consumers towards to a same product sometimes vary from different online sources. Then the question arises: what makes those reviews different and how different they are. This paper focuses on analyzing the discrepancy of online reviews on a product from different sources. It proposes a methodology to investigate the root cause of the discrepancy among the reviews from different sources using topic modeling based content analysis. Experimental results and analyses based on the data collected from popular e-commerce sites and online consumer communities are presented in this paper.