Nonparametric Bayesian Storyline Detection from Microtexts

TL;DR

This paper introduces a novel online non-parametric Bayesian method using dd-CRP for detecting evolving storylines in microtexts, effectively integrating temporal and linguistic data with efficient inference.

Contribution

It presents a new non-parametric Bayesian framework with a fixed-lag Gibbs sampling inference for storyline detection, improving efficiency and performance over heuristic methods.

Findings

Competitive with top TTG 2014 entries

Efficient linear-time inference achieved

Effective integration of temporal and linguistic info

Abstract

News events and social media are composed of evolving storylines, which capture public attention for a limited period of time. Identifying storylines requires integrating temporal and linguistic information, and prior work takes a largely heuristic approach. We present a novel online non-parametric Bayesian framework for storyline detection, using the distance-dependent Chinese Restaurant Process (dd-CRP). To ensure efficient linear-time inference, we employ a fixed-lag Gibbs sampling procedure, which is novel for the dd-CRP. We evaluate on the TREC Twitter Timeline Generation (TTG), obtaining encouraging results: despite using a weak baseline retrieval model, the dd-CRP story clustering method is competitive with the best entries in the 2014 TTG task.