SPEECH SEGMENTATION IN A SIMULATED BILINGUAL ENVIRONMENT: A CHALLENGE FOR STATISTICAL LEARNING?
SPEECH SEGMENTATION IN A SIMULATED BILINGUAL ENVIRONMENT: A CHALLENGE FOR STATISTICAL LEARNING?
复制标题
模拟双语环境中的语音分割：统计学习的挑战？
DOI：
10.1080/15475440802340101
复制
发表时间：
2009
期刊：
影响因子：
--
通讯作者：
Mitchel,AaronD
中科院分区：
文献类型：
--
作者：
Weiss,DanielJ;Gerfen,Chip;Mitchel,AaronD
使用人工语言流的研究表明，婴儿和成人都可以使用统计数据来正确地分词。然而，大多数研究只使用了一种输入语言。鉴于双语的盛行，多语言输入是如何分割的？如果学习者跨语言组合输入，可能会出现一个特殊的问题：重叠不同语言的特定单位的统计数据随后可能会改变并破坏正确的分割。我们的研究通过使用人工语言流来模拟成人l2学习者的早期分词阶段来解决这个问题。在四个实验中，参与者追踪了两种人工语言的多组统计数据。我们的研究结果表明，成人学习者可以同时跟踪两组统计数据，这表明他们在面对双语输入时形成了多重表征。这项工作，连同计划中的婴儿实验，揭示了双语研究中的一个核心问题，即确定当暴露于多种语言时，听众在什么时候可以形成多种表征。
Studies using artificial language streams indicate that infants and adults can use statistics to correctly segment words. However, most studies have utilized only a single input language. Given the prevalence of bilingualism, how is multiple language input segmented? One particular problem may occur if learners combine input across languages: The statistics of particular units that overlap different languages may subsequently change and disrupt correct segmentation. Our study addresses this issue by employing artificial language streams to simulate the earliest stages of segmentation in adult L2-learners. In four experiments, participants tracked multiple sets of statistics for two artificial languages. Our results demonstrate that adult learners can track two sets of statistics simultaneously, suggesting that they form multiple representations when confronted with bilingual input. This work, along with planned infant experiments, informs a central issue in bilingualism research, namely, determining at what point listeners can form multiple representations when exposed to multiple languages.