Hawkes Binomial Topic Model with Applications to Coupled Conflict-twitter Data

We consider the problem of modeling and clustering heterogeneous event data arising from coupled conflict event and social media data sets. In this setting conflict events trigger responses on social media and at the same time signals of grievance detected in social media may serve as leading indicators for subsequent conflict events. For this purpose we introduce the Hawkes Binomial Topic Model (HBTM) where marks, Tweets, and conflict event descriptions are represented as bags of words following a Binomial distribution. When viewed as a branching process, the daughter event bag of words is generated by randomly turning on/off parent words through independent Bernoulli random variables. We then use ExpectationMaximization to estimate the model parameters and branching structure of the process. The inferred branching structure is then used for Topic cascade detection, short-term forecasting, and investigating the causal dependence of grievance on social media and conflict events in recent elections in Nigeria and Kenya. 1. Background and Motivation. Twitter and other social media platforms have emerged as important tools for the public to communicate responses to crises and terrorist attacks [5] and more generally to communicate collectively, exchanging grievances that may catalyze mobilization [3, 24]. Research has focused on understanding public sentiment around these types of events and determining the central actors in the social network that are key to shaping public response [5] along with measuring short-term changes in the intensity of conflict using social media [29] or considering effects from regional instability [4]. At a more macro spatial-temporal scale, recent research has focused on modeling the endogenous and exogenous processes that generate thousands of terrorist and conflict events at the level of countries and years or decades. For this purpose point processes are used to model contagion effects in the risk of terrorist activity [21] and both contagion and exogenous rate fluctuations in conflict [17, 28, 27]. Because of the relative infrequency of conflict and terrorist events, having auxiliary data that can provide a signal for the risk of future events is highly desirable. While conflict events have been shown to influence overall regional instability [4], point process models of conflict to date have focused on univariate data [21, 17, 28, 27]. In this paper we use 1 imsart-aoas ver. 2014/10/16 file: main.tex date: August 6, 2017