Abstract

We present an approach to event coreference resolution by developing a general framework for clustering that uses supervised representation learning. We propose a neural network architecture with novel Clustering-Oriented Regularization (CORE) terms in the objective function. These terms encourage the model to create embeddings of event mentions that are amenable to clustering. We then use agglomerative clustering on these embeddings to build event coreference chains. For both within- and cross-document coreference on the ECB+ corpus, our model obtains better results than models that require significantly more pre-annotated information. This work provides insight and motivating results for a new general approach to solving coreference and clustering problems with representation learning.

Highlights

  • Event coreference resolution is the task of determining which event mentions expressed in language refer to the same real-world event instances

  • We propose a mechanism to modulate this training by introducing Clustering-Oriented Regularization (CORE) terms into the objective function of the learner; these terms impel the model to produce similar embeddings for coreferential event mentions, and dissimilar embeddings otherwise

  • We have presented a novel approach to event coreference resolution by combining supervised representation learning with non-parametric clustering

Read more

Summary

Introduction

Event coreference resolution is the task of determining which event mentions expressed in language refer to the same real-world event instances. We hypothesize that prior knowledge about the task itself can be usefully encoded into the representation learning objective For our task, this prior means that the embeddings of corefential event mentions should have similar embeddings to each other (a “natural clustering”, using the terminology of Bengio et al (2013)). Our model creates embeddings of event mentions that are directly conducive for the clustering task of building event coreference chains This is contrary to the indirect methods of previous work that rely on pairwise decision making followed by a separate model that aggregates the sometimes inconsistent decisions into clusters (Section 2). Our model’s improvements upon the baselines show that our supervised representation learning framework creates new embeddings that capture the abstract distributional relations between samples and their clusters, suggesting that our framework can be generalized to other clustering tasks

Related Work
Event Coreference Resolution Model
Model Overview
Supervised Representation Learning
Attractive Regularization
Repulsive Regularization
Loss Function
Agglomerative Clustering
Feature Extraction
Contextual
Document
Comparative
Experimental Design
Evaluation Measures
Models
Results
Conclusions and Future Work
Full Text
Paper version not known

Talk to us

Join us for a 30 min session where you can share your feedback and ask us any queries you have

Schedule a call

Disclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.