Scalable Detection of Sentiment-Based Contradictions

The analysis of user opinions expressed on the Web is becoming increasingly relevant to a variety of applications. It allows us to track the evolution of opinions or discussions in the blogosphere, or perform product surveys. The aggregation of sentiments and analysis of contradictions is another important application, which becomes effective since we are able to capture the diversity in sentiments on different topics with more precision and on a large scale. Though, there is still a need for a scalable way of sentiment aggregation with respect to the time dimension, which preserves enough information to capture contradictions. In this paper, we are focusing on the problem of finding sentimentbased contradictions at a large scale. First, we define two types of contradictions, depending on the distributions of opposite sentiments over time. Second, we introduce a novel measure of contradiction based on the mean value and the variance of sentiments among different texts. Third, we propose a scalable method for identifying both types of contradictions at different time scales. We evaluate the performance of our method using synthetic and realworld datasets, as well as a user-study. The experiments demonstrate the effectiveness of the proposed method in capturing contradictions in a scalable manner.