Supervised Model Learning with Feature Grouping based on a Discrete Constraint
Jun Suzuki and Masaaki Nagata
The 51st Annual Meeting of the Association for Computational Linguistics - Short Papers (ACL Short Papers 2013)
Sofia, Bulgaria, August 4-9, 2013
This paper proposes a framework of supervised model learning that simultaneously achieves a feature grouping concept for obtaining lower complexity models. The main idea of our method is simply incorporating a discrete constraint during the model learning with the help of the dual decomposition technique. Experiments on well-studied two NLP tasks, namely, dependency parsing and NER demonstrate that our method can provide the state-ofthe-art performance with parameter’s degree of freedom is surprisingly small, i.e., less than hundred. This significant reduction enables us to provide a compact model representation for decoding phrase especially useful for real use.
Conference Manager (V2.61.0 - Rev. 2792M)