Decoding continuous speech in infancy

解码婴儿期的连续语音

• 批准号: 355894-2013
• 负责人: Johnson, Elizabeth
• 金额: $ 2.4万
• 依托单位: University of Toronto
• 依托单位国家: 加拿大
• 项目类别: Discovery Grants Program - Individual
• 财政年份: 2015
• 资助国家: 加拿大
• 起止时间: 2015-01-01 至 2016-12-31
• 项目状态: 已结题
• 来源: https://www.nserc-crsng.gc.ca/ase-oro/Details-Detailles_eng.asp?id=572687
• 关键词: Decoding; continuous; speech; infancy

The first aim of this project is to examine how children begin to extract word forms from continuous speech. The second aim is to examine how children link extracted word forms to intended word meanings. Both tasks are far from trivial. Finding words in speech is is a challenge because natural speech contains no fully reliable cues to word boundaries. Adults only perceive speech as a string of discrete words because they know language-specific cues to word boundaries. Attaching meaning to word forms is also difficult because even in overt labelling scenarios, there is rarely a one-to-one unambiguous mapping between an object and its intended label. To appreciate the difficulty of finding words in speech and determining their intended meaning, imagine a child who hears her mother enthusiastically utter the word 'pigeon' while walking along a city sidewalk. What tasks must be achieved before the child can add this word to their vocabulary? First, the child must determine whether the utterance contains one word ('pigeon') or two words ('pih gin)'. Assuming the child can work out that 'pigeon' is one word, how does she then determine what the word means? Does it refer to the cars whizzing past, the stop sign on the corner, her loose shoelace, or the gait of the avian creature her mom is pointing at? Many contemporary models of language acquisition and speech perception attribute children's success in these word learning situations to children's keen ability to track and compute statistical relationships between events and objects in their environment. For example, the child might be able to determine that the utterance 'pidgin' is one word rather than two because the syllables 'pih' and 'gin' consistently co-occur. The child might also realize that this one word refers to a particular type of bird because she notices that nearly every time the word is uttered, a grey bird is in sight. In the current project, we explore the ecological validity of these sorts of distributional models of language acquisition. By uncovering the mechanisms involved in normal language development, we will position future researchers to better understand both normal and abnormal patterns of speech and language development.

这个项目的第一个目的是研究儿童是如何开始从连续的语音中提取单词形式的。第二个目的是研究儿童如何将提取的单词形式与预期的单词含义联系起来。这两项任务都远非微不足道。在语音中找到单词是一个挑战，因为自然语音不包含完全可靠的单词边界线索。成年人只把言语看作是一串离散的单词，因为他们知道单词边界的特定语言线索。将意义附加到单词形式也很困难，因为即使在公开的标签场景中，对象和其预期标签之间也很少有一对一的明确映射。为了理解在言语中寻找单词并确定其意图的困难，想象一下一个孩子在沿着城市人行道行走时听到她的母亲热情地说出“鸽子”这个词。在孩子能把这个词添加到他们的词汇表之前，必须完成什么任务？首先，孩子必须确定话语是否包含一个词（“鸽子”）或两个词（“pih gin”）。假设孩子能理解“鸽子”是一个词，那么她如何确定这个词的意思呢？它指的是呼啸而过的汽车，街角的停车标志，她松了的鞋带，还是她妈妈指着的鸟类生物的步态？许多当代语言习得和言语感知模型将儿童在这些单词学习情境中的成功归因于儿童敏锐的跟踪和计算其环境中的事件和对象之间的统计关系的能力。例如，儿童可能能够确定话语“洋泾浜”是一个词而不是两个词，因为音节“pih”和“gin”一致地共同出现。孩子也可能意识到这个词指的是一种特定类型的鸟，因为她注意到几乎每次说出这个词时，都会看到一只灰色的鸟。 在当前的项目中，我们将探索这些语言习得分布模型的生态有效性。通过揭示正常语言发展中所涉及的机制，我们将使未来的研究人员能够更好地理解正常和异常的言语和语言发展模式。