Structure Compilation: Trading Structure for Features
Abstract
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.
Cite
Text
Liang et al. "Structure Compilation: Trading Structure for Features." International Conference on Machine Learning, 2008. doi:10.1145/1390156.1390231Markdown
[Liang et al. "Structure Compilation: Trading Structure for Features." International Conference on Machine Learning, 2008.](https://mlanthology.org/icml/2008/liang2008icml-structure/) doi:10.1145/1390156.1390231BibTeX
@inproceedings{liang2008icml-structure,
title = {{Structure Compilation: Trading Structure for Features}},
author = {Liang, Percy and Iii, Hal Daumé and Klein, Dan},
booktitle = {International Conference on Machine Learning},
year = {2008},
pages = {592-599},
doi = {10.1145/1390156.1390231},
url = {https://mlanthology.org/icml/2008/liang2008icml-structure/}
}