Combining Temporal Information and Topic Modeling for Cross-Document Event Ordering

TL;DR

This paper presents a method combining temporal information and topic modeling to improve cross-document event coreference resolution and event ordering in news articles, achieving state-of-the-art results.

Contribution

It introduces a novel approach integrating temporal cues and semantic topic modeling for event coreference resolution across documents.

Findings

Achieved highest Micro-average F-score in SemEval2015 Task 4

Improved event ordering accuracy by up to 6%

Identified limitations of topic modeling in this context

Abstract

Building unified timelines from a collection of written news articles requires cross-document event coreference resolution and temporal relation extraction. In this paper we present an approach event coreference resolution according to: a) similar temporal information, and b) similar semantic arguments. Temporal information is detected using an automatic temporal information system (TIPSem), while semantic information is represented by means of LDA Topic Modeling. The evaluation of our approach shows that it obtains the highest Micro-average F-score results in the SemEval2015 Task 4: TimeLine: Cross-Document Event Ordering (25.36\% for TrackB, 23.15\% for SubtrackB), with an improvement of up to 6\% in comparison to the other systems. However, our experiment also showed some draw-backs in the Topic Modeling approach that degrades performance of the system.