Logistic Normal Priors for Unsupervised Probabilistic Grammar Induction
Logistic Normal Priors for Unsupervised Probabilistic Grammar Induction
复制标题
DOI：
--
复制
发表时间：
2008-12
期刊：
影响因子：
--
通讯作者：
Shay B. Cohen;Kevin Gimpel;Noah A. Smith
中科院分区：
文献类型：
--
作者：
Shay B. Cohen;Kevin Gimpel;Noah A. Smith
我们探索了一种新的概率文法的贝叶斯模型，这是一个离散结构上的分布族，包括隐马尔可夫模型和概率上下文无关文法。我们的模型将相关主题模型框架扩展到概率语法，利用逻辑正态分布作为语法参数的先验。我们推导了该模型的变分EM算法，并对自然语言依存分析的无监督文法归纳任务进行了实验。我们表明，我们的模型比以前使用不同先验的模型取得了更好的结果。
We explore a new Bayesian model for probabilistic grammars, a family of distributions over discrete structures that includes hidden Markov models and probabilistic context-free grammars. Our model extends the correlated topic model framework to probabilistic grammars, exploiting the logistic normal distribution as a prior over the grammar parameters. We derive a variational EM algorithm for that model, and then experiment with the task of unsupervised grammar induction for natural language dependency parsing. We show that our model achieves superior results over previous models that use different priors.