Distributional Learning of Speech Sounds: An Exploratory Study Into the Effects of Prior Language Experience,Language Learning

当前位置： X-MOL 学术 › Lang. Learn. › 论文详情

Our official English website, www.x-mol.net, welcomes your feedback! (Note: you will need to create a separate account there.)

Distributional Learning of Speech Sounds: An Exploratory Study Into the Effects of Prior Language Experience
Language Learning ( IF 3.5 ) Pub Date : 2020-09-09 , DOI: 10.1111/lang.12432
Kateřina Chládková _{1,

2} , Šárka Šimáčková ₃

Affiliation

Distributional learning is typically understood as (unattended) tracking of stimulus probabilities. Distributional training with speech yields mixed results and the influencing factors have not yet been fully investigated. This study explored whether prior linguistic experience could have an effect on distributional learning outcomes. Czech and Greek adults, whose native languages contain and lack abstract length categories, respectively, were exposed to novel vowels falling into unimodal or bimodal distributions along the durational dimension. A trending interaction suggested that the Czechs and the Greeks might have been affected differently by the distributional exposure. Improved discrimination of the “trained” contrast was observed in bimodally exposed Czechs (whose prior expectations about length categories could guide learning) and, rather surprisingly, in unimodally exposed Greeks (who, lacking any expectations, might have listened in a noncategorical, auditory mode). Prior linguistic experience could thus affect whether and how experienced language users exploit new distributional speech statistics. This proposal needs to be assessed in future studies.

中文翻译：

语音的分布式学习：先验语言体验影响的探索性研究

分布式学习通常被理解为对刺激概率的（无人值守）跟踪。具有语音的分布式训练产生的结果参差不齐，影响因素尚未得到充分研究。这项研究探讨了以前的语言经验是否会对分配学习成果产生影响。捷克人和希腊人的母语分别包含和缺少抽象的长度类别，他们接触的新颖元音沿时空方向呈单峰或双峰分布。趋势相互作用表明，捷克人和希腊人可能受到分布风险的不同影响。在暴露于双峰状态的捷克人中，“训练过的”对比度得到了更好的区分（他们先前对长度类别的期望可以指导学习），并且令人惊讶的是，在暴露于单峰状态的希腊语中（没有任何期望的人，可能会以非分类的听觉方式进行聆听）。因此，先前的语言经验可能会影响经验丰富的语言用户是否以及如何利用新的分布式语音统计数据。该建议需要在以后的研究中进行评估。

更新日期：2020-09-09

点击分享查看原文

点击收藏

公开下载

阅读更多本刊最新论文