Large Margin Methods for Structured and Interdependent Output Variables
The Journal of Machine Learning Research
EMNLP '02 Proceedings of the ACL-02 conference on Empirical methods in natural language processing - Volume 10
Online large-margin training of dependency parsers
ACL '05 Proceedings of the 43rd Annual Meeting on Association for Computational Linguistics
Learning structural SVMs with latent variables
ICML '09 Proceedings of the 26th Annual International Conference on Machine Learning
Multilingual dependency analysis with a two-stage discriminative parser
CoNLL-X '06 Proceedings of the Tenth Conference on Computational Natural Language Learning
Learning from partially annotated sequences
ECML PKDD'11 Proceedings of the 2011 European conference on Machine learning and knowledge discovery in databases - Volume Part I
Stanford's multi-pass sieve coreference resolution system at the CoNLL-2011 shared task
CONLL Shared Task '11 Proceedings of the Fifteenth Conference on Computational Natural Language Learning: Shared Task
Rule and tree ensembles for unrestricted coreference resolution
CONLL Shared Task '11 Proceedings of the Fifteenth Conference on Computational Natural Language Learning: Shared Task
Entropy-Guided feature generation for structured learning of portuguese dependency parsing
PROPOR'12 Proceedings of the 10th international conference on Computational Processing of the Portuguese Language
CoNLL-2012 shared task: Modeling Multilingual Unrestricted Coreference in OntoNotes
CoNLL '12 Joint Conference on EMNLP and CoNLL - Shared Task
Deterministic coreference resolution based on entity-centric, precision-ranked rules
Computational Linguistics
Hi-index | 0.00 |
We describe a machine learning system based on large margin structure perceptron for unrestricted coreference resolution that introduces two key modeling techniques: latent coreference trees and entropy guided feature induction. The proposed latent tree modeling turns the learning problem computationally feasible. Additionally, using an automatic feature induction method, we are able to efficiently build nonlinear models and, hence, achieve high performances with a linear learning algorithm. Our system is evaluated on the CoNLL-2012 Shared Task closed track, which comprises three languages: Arabic, Chinese and English. We apply the same system to all languages, except for minor adaptations on some language dependent features, like static lists of pronouns. Our system achieves an official score of 58.69, the best one among all the competitors.