The learner as statistician: three principles of computational success in language acquisition.
The learner as statistician: three principles of computational success in language acquisition.
复制标题
作为统计学家的学习者：语言习得中计算成功的三个原则。
DOI：
10.1111/j.1467-7687.2009.00827.x
复制
发表时间：
2009
影响因子：
3.7
通讯作者：
Morgan,James
中科院分区：
文献类型：
--
作者：
Soderstrom,Melanie;Conwell,Erin;Feldman,Naomi;Morgan,James
统计学习是语言习得的新范式。仔细阅读最近的会议计划或期刊内容，就会发现许多倡导或批评统计学习的工作。语言习得将继续受益于各种理论和方法，但是，正如本期文章所举的例子，统计学习已经从次要角色发展到核心角色。为了确保持久的影响，统计方法现在必须从零碎的演示转向语言学习的一般理论。统计学习与其成功的主流范式形成鲜明对比。原理和参数方法（乔姆斯基，1981）假设丰富的先天禀赋，有限的处理能力和贫乏的输入，而统计学习假设输入丰富，学习者具有足够的计算能力来提取相关的语言模式。统计学习模型之所以有吸引力，是因为原则上它们采用了强大的通用任务机制来解决语言习得的难题。此外，对成人和婴儿的行为研究表明，人类在类似语言的任务中使用统计学习（Gómez, 2002; Goodsitt, Morgan & Kuhl, 1992; Maye, Werker & Gerken, 2002; Saffran, Aslin & Newport, 1996; Saffran, Newport & Aslin， 1996等）。
Statistical learning is the new paradigm of language acquisition. A perusal of recent conference programs or journal contents reveals much work advocating–or criticizing–statistical learning. Language acquisition will continue to benefit from a variety of theories and methods, but, as the articles in this issue exemplify, statistical learning has progressed from being a minor player towards a central role. To ensure a lasting impact, statistical approaches must now move from piecemeal demonstrations towards a general theory of language learning.Statistical learning stands in contrast to the predominant paradigm that it succeeded. The principles and parameters approach (Chomsky, 1981) assumed rich innate endowment, limited processing abilities and impoverished input, whereas statistical learning assumes that input is rich and that learners possess sufficient computational sophistication to extract relevant linguistic patterns. Statistical learning models are attractive because in principle they recruit powerful, task-general machinery to solve difficult problems of language acquisition. Furthermore, behavioural findings with both adults and infants suggest that humans use statistical learning in language-like tasks (Gómez, 2002; Goodsitt, Morgan & Kuhl, 1992; Maye, Werker & Gerken, 2002; Saffran, Aslin & Newport, 1996, Saffran, Newport & Aslin, 1996, inter alia).