Nonparametric Bayesian Storyline Detection from Microtexts
This addresses the challenge of identifying storylines for applications in news and social media analysis, but it is incremental as it builds on existing Bayesian methods with a novel inference procedure.
The paper tackled the problem of detecting evolving storylines from microtexts like news and social media by proposing an online non-parametric Bayesian framework using the distance-dependent Chinese Restaurant Process (dd-CRP). The result showed that, despite using a weak baseline retrieval model, the method was competitive with the best entries in the 2014 TREC Twitter Timeline Generation task.
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.