Structure compilation
Percy Liang, Hal Daumé, Dan Klein · International Conference on Machine Learning · 2008
Structured models often achieve excellent performance but can be slow at test time. We investigate structure compilation, where we replace structure with features, which are often computationally simpler but unfortunately statistically more complex. We analyze this tradeoff theoretically and empirically on three natural language processing tasks. We also introduce a simple method to transfer predictive power from structure to features via unlabeled data, while incurring a minimal statistical penalty.
Cited by 69 works. doi:10.1145/1390156.1390231
Authors
References
- Conditional Random Fields: Probabilistic Models for Segmenting and Labeling Sequence Data
- Convergence of Stochastic Processes.
- Model compression
- Piecewise Training for Undirected Models
- Information Theory and Statistics: A Tutorial
- Learning from Multiple Sources
- Extracting comprehensible models from trained neural networks
- Learning and inference over constrained output
- Discriminative Log-Linear Grammars with Latent Variables
- Parameter Estimation For Statistical Parsing Models: Theory And Practice Of Distribution-Free Methods