Learning the Structure of Variable-Order CRFs: a finite-state perspective - Archive ouverte HAL Accéder directement au contenu
Communication Dans Un Congrès Année : 2017

Learning the Structure of Variable-Order CRFs: a finite-state perspective

Résumé

The computational complexity of linear-chain Conditional Random Fields (CRFs) makes it difficult to deal with very large label sets and long range dependencies. Such situations are not rare and arise when dealing with morphologically rich languages or joint labelling tasks. We extend here recent proposals to consider variable order CRFs. Using an effective finite-state representation of variable-length dependencies, we propose new ways to perform feature selection at large scale and report experimental results where we outperform strong baselines on a tagging task.
Fichier principal
Vignette du fichier
D17-1044.pdf (178.23 Ko) Télécharger le fichier
Origine : Fichiers éditeurs autorisés sur une archive ouverte
Loading...

Dates et versions

hal-01710793 , version 1 (20-08-2019)

Identifiants

Citer

Thomas Lavergne, François Yvon. Learning the Structure of Variable-Order CRFs: a finite-state perspective. Conference on Empirical Methods in Natural Language Processing, Sep 2017, Copenhagen, Denmark. pp.433 - 439, ⟨10.18653/v1/D17-1044⟩. ⟨hal-01710793⟩
58 Consultations
41 Téléchargements

Altmetric

Partager

Gmail Facebook X LinkedIn More