Grammars leak: Modeling how phonotactic generalizations interact within the grammar
Grammars leak: Modeling how phonotactic generalizations interact within the grammar
复制标题
语法泄漏：建模音位概括如何在语法中相互作用
DOI：
10.1353/lan.2011.0096
复制
发表时间：
2011
期刊：
影响因子：
2.1
通讯作者：
Andrew Martin
中科院分区：
文献类型：
--
作者：
Andrew Martin
我提供了来自纳瓦霍语和英语的证据，证明语素内部音位制约的较弱、梯度版本，如禁止英语中的双辅音，即使跨越韵律单词的界限也是有效的。我认为，这些词汇偏向是最大熵语音定向学习算法的结果，该算法最大化了学习数据的概率，但也包含了一个惩罚复杂语法的平滑术语。当学习者试图构建一种语法，其中的一些约束是对形态结构视而不见的，它低估了违反语素内部音位的复合词的频率。我进一步展示了，随着时间的推移，这种学习偏见可能会导致纳瓦霍语和英语中出现的词汇偏见。
I present evidence from Navajo and English that weaker, gradient versions of morpheme-internal phonotactic constraints, such as the ban on geminate consonants in English, hold even across prosodic word boundaries. I argue that these lexical biases are the result of a maximum entropy phonotactic learning algorithm that maximizes the probability of the learning data, but that also contains a smoothing term that penalizes complex grammars. When this learner attempts to construct a grammar in which some constraints are blind to morphological structure, it underpredicts the frequency of compounds that violate a morpheme-internal phonotactic. I further show how, over time, this learning bias could plausibly lead to the lexical biases seen in Navajo and English.