Doctoral Dissertation Research: Units of orthographic complexity and their influence in speech production

博士论文研究：拼写复杂性单位及其对语音产生的影响

• 批准号: 1844429
• 负责人: Thomas Bever
• 金额: $ 1.47万
• 依托单位: University of Arizona
• 依托单位国家: 美国
• 项目类别: Standard Grant
• 财政年份: 2019
• 资助国家: 美国
• 起止时间: 2019-03-01 至 2021-02-28
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=1844429&HistoricalAwards=false
• 关键词: Doctoral; Dissertation; Research; Units; orthographic

How does literacy impact speech? There is evidence that the number of symbols in the written form of a sound can influence how long that sound is produced, even when a written form is not explicitly present. For example, the final /k/ sounds in the English words "tic", "click", and "clique" are represented by 1, 2, and 3 letters, respectively. When people produce these words, the final /k/ of "click" is produced longer than than the final /k/ of "tic". And the final /k/ of "clique" is produced longer still. Therefore, it has been hypothesized that as the complexity of a written form increases, so too does the speech duration of the associated sound, all else being equal. This phenomenon is inconsistent with some models of language production which argue that written information should not influence speech, especially when written information is not explicitly present. Additionally, it remains unclear exactly which types of written information can influence speech duration as there is disagreement about the mechanisms that drive this effect. Some argue the relationship between sounds and written symbols underlie this effect. Others argue the amount of visual information itself regardless of any connection to sounds is driving these duration differences. This study examines the relationship between sounds, symbols, and speech duration to develop our understanding of what types of written information can alter speech patterns. Addressing this question will inform theoretical models of human speech production. Additionally, data collected through this research can be incorporated into computational speech synthesis and speech recognition systems for more accurate, robust, and natural models of human speech. A challenge of previous research in this area was that it relied upon natural speech data which includes many external factors across individuals that can influence speech duration outside of written forms (e.g. word frequency). To gain greater control of these variables, the current study will employ a three-day novel-word-learning experiment in which participants learn sets of novel non-word homophones along with one of several novel writing systems. Novel homophones in a set are learned as the same word phonologically but are written distinctly. Written complexity for homophones may vary by: a) number of whole characters; b) number of pen strokes; c) number of components with phonological associations; d) number of components without phonological associations. Participants will produce learned words in a picture-naming task without written forms explicitly present. Participants' production will be recorded and speech duration will be measured. Duration among homophones in a set and across writing systems will be compared to determine: a) if increased written complexity influences duration even when written forms are not explicitly present; b) which types of written information may influence duration and whether written complexity must have a phonological association to influence speech; and c) at which point in the training process speech patterns change and whether these patterns persist through the end of the experimental period. This data will comprise a portion of the doctoral dissertation. An earlier portion of the dissertation investigated similar durational phenomenon in Japanese speakers in regards to written character complexity. To remain consistent with previous data set, retain the ability to compare results from the current study with earlier data in the dissertation, and provide the opportunity to replicate previous findings in the current study, participants for the current study will be native Japanese speakers.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.

识字能力对语言有何影响？有证据表明，即使在没有明确的书面形式的情况下，声音书面形式中的符号数量也会影响声音产生的时长。例如，英语单词“tic”、“Click”和“CLIQUE”中最后的/k/发音分别由1、2和3个字母表示。当人们说出这些单词时，“Click”的尾音/k/比“tic”的尾音/k/要长。而“CLIQUE”的尾音/k/产生的时间更长。因此，假设在其他条件相同的情况下，随着书面形式的复杂性增加，相关声音的语音时长也会增加。这一现象与一些语言产生模型不一致，这些模型认为书面信息不应该影响言语，特别是当书面信息不是明确存在的时候。此外，目前还不清楚哪种类型的书面信息会影响语音时长，因为人们对这种影响的机制存在分歧。一些人认为，声音和文字符号之间的关系构成了这种效果的基础。其他人认为，视觉信息本身的数量，无论与声音有任何联系，都是导致这些时长差异的原因。这项研究考察了声音、符号和言语时长之间的关系，以加深我们对哪些类型的书面信息可以改变言语模式的理解。解决这个问题将为人类言语产生的理论模型提供信息。此外，通过这项研究收集的数据可以被合并到计算语音合成和语音识别系统中，以实现更准确、更稳健和更自然的人类语音模型。这一领域以往研究的一个挑战是，它依赖于自然语音数据，其中包括许多个体之间的外部因素，这些因素可以影响书面形式以外的语音持续时间(例如，词频)。为了更好地控制这些变量，目前的研究将采用为期三天的小说单词学习实验，参与者在学习几种小说写作系统之一的同时，学习几组新的非单词同音异义词。一组新奇的同音字在语音上是作为同一个词学习的，但在书写上是不同的。同音字的书写复杂性可以由以下因素变化：a)整个字符的数量；b)笔划的数量；c)具有语音关联的部件的数量；d)没有语音关联的部件的数量。参与者将在没有明确书面形式的情况下，在图片命名任务中产生所学的单词。参与者的发言将被记录下来，并将测量演讲时长。将比较一组和跨书写系统中的同音字之间的持续时间，以确定：a)即使没有明确的书写形式，增加的书写复杂性是否会影响持续时间；b)哪些类型的书写信息可能影响持续时间，以及书写的复杂性是否必须具有音系学关联才能影响语音；以及c)在训练过程中的哪个时间点，语音模式改变，以及这些模式是否持续到实验期结束。这些数据将构成博士论文的一部分。论文的前一部分研究了说日语的人在书写字符复杂性方面的类似持续时间现象。为了与以前的数据集保持一致，保留将当前研究的结果与论文中的早期数据进行比较的能力，并提供复制当前研究中以前的发现的机会，当前研究的参与者将是以日语为母语的人。该奖项反映了NSF的法定使命，并通过使用基金会的智力优势和更广泛的影响审查标准进行评估，被认为值得支持。