Learning with Lookahead: Can History-Based Models Rival Globally Optimized Models?
Learning with Lookahead: Can History-Based Models Rival Globally Optimized Models?
复制标题
DOI：
--
复制
发表时间：
2011-06
期刊：
影响因子：
--
通讯作者：
Yoshimasa Tsuruoka;Yusuke Miyao;Jun'ichi Kazama
中科院分区：
文献类型：
--
作者：
Yoshimasa Tsuruoka;Yusuke Miyao;Jun'ichi Kazama
本文表明，基于历史的模型的性能，可以显着提高在状态空间进行前瞻时，每个分类决策。我们不是简单地使用分类器输出的最佳动作，而是通过查看未来动作的可能序列并评估这些动作序列实现的最终状态来确定最佳动作。我们提出了一个基于感知器的参数优化方法，这个学习框架，并显示其收敛性能。建议的框架进行评估的词性标注，组块，命名实体识别和依赖分析，使用标准的数据集和功能。实验结果表明，基于历史的模型与前瞻的竞争力的全局优化模型，包括条件随机场（CRF）和结构化感知器。
This paper shows that the performance of history-based models can be significantly improved by performing lookahead in the state space when making each classification decision. Instead of simply using the best action output by the classifier, we determine the best action by looking into possible sequences of future actions and evaluating the final states realized by those action sequences. We present a perceptron-based parameter optimization method for this learning framework and show its convergence properties. The proposed framework is evaluated on part-of-speech tagging, chunking, named entity recognition and dependency parsing, using standard data sets and features. Experimental results demonstrate that history-based models with lookahead are as competitive as globally optimized models including conditional random fields (CRFs) and structured perceptrons.