Automatic Story Construction from News Articles in an Online Fashion

TL;DR

This paper introduces an online story construction system that tracks news story evolution using a novel sliding window approach and real-time community detection, enabling efficient, incremental story building with visualizations.

Contribution

It proposes the Inching Window method and an on-the-fly Louvain-based community detection for real-time news story tracking, a novel approach in online news analysis.

Findings

Successful step-by-step news story construction

Effective real-time community detection in news streams

Visualizations support story evolution understanding

Abstract

This paper presents a novel story construction system to track the evolution of stories in an online fashion. The proposed system uses a novel sliding window solution, named Inching Window, allowing the processing of each new data element on-the-fly. To assign a new data element into a community in a fast and memory-efficient manner, we apply the modularity maximization idea of Louvain method on-the-fly. As part of the experimental validation, we provide step by step construction of a meaningful news story and support the case with a set of visualizations.