Timezone: »

 
Spotlight
Discriminative Log-Linear Grammars with Latent Variables
Slav Petrov · Dan Klein

Mon Dec 03 08:10 PM -- 08:25 PM (PST) @ None

We demonstrate that log-linear grammars with latent variables can be practically trained using discriminative methods. Central to efficient discriminative training is a hierarchical pruning procedure which allows feature expectations to be efficiently approximated. We investigate an EM alternating procedure as well as a direct, gradient-based procedure and show that direct optimization is superior. On full-scale treebank parsing experiments, the discriminative latent models outperform both the comparable generative latent models as well as the discriminative non-latent baselines.

Author Information

Slav Petrov (Google)
Dan Klein (UC Berkeley)

Related Events (a corresponding poster, oral, or spotlight)

More from the Same Authors