Translating Granularity of Event Slots into Features for Event Coreference Resolution.

Abstract

Using clues from event semantics to solve coreference, we present an "event template" approach to cross-document event coreference resolution on news articles. The approach uses a pairwise model, in which event information is compared along five semantically motivated slots of an event template. The templates, filled in on the sentence level for every event mention from the data set, are used for supervised classification. In this study, we determine granularity of events and we use the grain size as a clue for solving event coreference. We experiment with a newly-created granularity ontology employing granularity levels of locations, times and human participants as well as event durations as features in event coreference resolution. The granularity ontology is available for research. Results show that determining granularity along semantic event slots, even on the sentence level exclusively, improves precision and solves event coreference with scores comparable to those achieved in related work.

Resolving Event Coreference with Supervised Representation Learning and Clustering-Oriented Regularization

“…Our model obtains strong results on withinand cross-document event coreference resolution, matching or outperforming the system of Cybulska and Vossen (2015) on the ECB+ corpus on all six evaluation measures. We achieve these gains despite the fact that our model requires significantly less pre-annotated or pre-detected information in terms of the internal event structure.…”

“…Previous work on model design for event coreference has focused on clustering over a linguistically rich set of features. Most models require a pairwise-prediction based supervised learning step which predicts whether or not a pair of event mentions is coreferential (Bagga and Baldwin, 1999;Cybulska and Vossen, 2015). Other work focuses on the clustering step itself, aggregating local pairwise decisions into clusters, for example by graph partitioning ).…”

Precision Event Coreference Resolution Using Neural Network Classifiers

“…Works specific to WD event coreference include pairwise classifiers (Ahn, 2006;), graph-based clustering ) and information propagation (Liu et al, 2014). Works focusing purely on CD coreference include Cybulska and Vossen (2015b) who created pairwise classifiers using features indicating granularities of event slots, and in another work (Cybulska and Vossen, 2015a), use discourse analysis at the document level along with 'sentence' templates amongst documents that have possibly coreferent events. Several papers have studied event extraction and event coreference as a joint process (Araki and Mitamura, 2015;Lu and Ng, 2017).…”

Event Linking with Sentential Features from Convolutional Neural Networks

“…A variation of the eventcoreference resolution task extends the scope to cross-document relations. Cybulska and Vossen (2015) approach this task with various classification models and propose to use a type-specific granularity hierarchy for feature values. Lee et al (2012) further extend the task definition by jointly resolving entity and event coreference, through several iterations of mention-cluster merge operations.…”

“…Event features from literature So far, a wide range of features has been used for the representation of events and relations for extraction (Zhou et al, 2005;Mintz et al, 2009;Sun et al, 2011;Krause et al, 2015) and coreference resolution (Bejan and Harabagiu, 2010;Lee et al, 2012;Araki and Mitamura, 2015;Cybulska and Vossen, 2015) purposes. The following is an attempt to list the most common classes among them, along with examples:…”