Event Coreference Resolution using Mincut based Graph Clustering

Abstract

ABSTRACT

Chinese Event Coreference Resolution: An Unsupervised Probabilistic Model Rivaling Supervised Resolvers

“…Improvements to this standard approach include the use of feature weighting to train a better model (McConky et al, 2012), and graph-based clustering algorithms to produce event coreference clusters (Chen and Ji, 2009;Sangeetha and Arock, 2012). Chen et al (2011) train multiple classifiers to handle coreference between event mentions of different syntactic types (e.g., verb-noun coreference, noun-noun coreference) on the OntoNotes corpus (Pradhan et al, 2007).…”

Proceedings of The 20th SIGNLL Conference on Computational Natural Language Learning

“…Previous approaches for modeling event mentions in context of coreference resolution (Bejan and Harabagiu, 2010;Sangeetha and Arock, 2012;) make either use of external feature sources with limited cross-domain availability like WordNet (Fellbaum, 1998) and FrameNet (Baker et al, 1998), or show low performance. At the same time, recent literature proposes a new kind of feature class for modeling events (and relations) in order to detect mentions and extract their arguments, i.e., sentential features from event-/relationmention representations that have been created by taking the full extent and surrounding sentence of a mention into account .…”

Joint Learning for Event Coreference Resolution

“…Existing event coreference resolvers have been evaluated on different corpora, such as MUC (e.g., Humphreys et al (1997)), ACE (e.g., Ahn (2006), Ji (2009), McConky et al (2012), Sangeetha and Arock (2012), Ng (2015, 2016), Krause et al (2016)), OntoNotes (e.g., Chen et al (2011)), the Intelligence Community corpus (e.g., Cybulska and Vossen (2012), , ), the ECB corpus (e.g., Lee et al (2012), Bejan and Harabagiu (2014)) and its extension ECB+ (e.g., Yang et al (2015)), and ProcessBank (e.g., Araki and Mitamura (2015)). The newest event coreference corpora are the ones used in the KBP 2015 and 2016 Event Nugget Detection and Coreference shared tasks, in which the best performers in 2015 and 2016 are RPI's system (Hong et al, 2015) and UTD's system , respectively.…”