SenseLens: An Efficient Social Signal Conditioning System for True Event Detection

Abstract

This article narrows the gap between physical sensing systems that measure physical signals and social sensing systems that measure information signals by (i) defining a novel algorithm for extracting information signals (building on results from text embedding) and (ii) showing that it increases the accuracy of truth discovery—the separation of true information from false/manipulated one. The work is applied in the context of separating true and false facts on social media, such as Twitter and Reddit, where users post predominantly short microblogs. The new algorithm decides how to aggregate the signal across words in the microblog for purposes of clustering the miscroblogs in the latent information signal space, where it is easier to separate true and false posts. Although previous literature extensively studied the problem of short text embedding/representation, this article improves previous work in three important respects: (1) Our work constitutes unsupervised truth discovery, requiring no labeled input or prior training. (2) We propose a new distance metric for efficient short text similarity estimation, we call Semantic Subset Matching , that improves our ability to meaningfully cluster microblog posts in the latent information signal space. (3) We introduce an iterative framework that jointly improves miscroblog clustering and truth discovery. The evaluation shows that the approach improves the accuracy of truth-discovery by 6.3%, 2.5%, and 3.8% (constituting a 38.9%, 14.2%, and 18.7% reduction in error, respectively) in three real Twitter data traces.