SEGMENTAL AND PROSODIC ASPECTS OF SPEECH PLANNING

言语规划的片段和韵律方面

• 批准号: 2014520
• 负责人: STEFANIE SHATTUCK-HUFNAGEL
• 金额: $ 7.75万
• 依托单位: MASSACHUSETTS INSTITUTE OF TECHNOLOGY
• 依托单位国家: 美国
• 财政年份: 1993
• 资助国家: 美国
• 起止时间: 1993-12-01 至 2000-11-30
• 项目状态: 已结题
• 来源: https://reporter.nih.gov/project-details/2014520
• 关键词: auditory discrimination; behavioral /social science research tag; computer system design /evaluation; information systems; model; neural information processing; psycholinguistics; semantics; sound; sound frequency; speech; speech recognition; vocalization

The research extends existing models of the segmental and prosopic planning processes and evaluates an integrated model of both of these aspects of the phonological planning process. Competing models of segmental planning, particularly models of the serial ordering of individual phonemic segments, are evaluated using data form existing corpora of errors from spontaneous speech, supplemented by additional examples to be collected. Models of prosopic planning, particularly the placement of intonational prominences (pitch accents), are evaluated in experimentally-elicited text sentences as well as in a database of prosopically-labelled FM radio speech. Finally, a model of the integration of segmental and prosopic planning is tested with data from prosopic analysis of segmental errors, made possible by collection of a new corpus of tape-recorded errors from spontaneous speech, and from experimentally elicited utterances designed to contrast the effects of syntactic vs. prosopic structure on error patterns. In the course of the work, existing databases of speech errors and of prosopically labelled speech will be enlarged and reorganized in ways that will facilitate their distribution to the research community. A more thoroughgoing understanding of these phonological aspects of the speech production planning process, an their relation both to syntactic/semantic planning on the one hand and to articulatory motor control on the other, will provide advantages in several areas, including an improved theoretical basis for the development of therapeutic treatments for disadvantaged speakers (including the hearing impaired and patients with certain kind of aphasia), a stronger foundation for advances in automatic speech recognition and synthesis that will aid in development of aids for the language and speech handicapped, and a better-grounded approach to effective ways of teaching the prosody of English as a second language.

该研究扩展了现有的分段和前景模型 规划流程并评估这两者的集成模型 语音规划过程的各个方面。 竞争型号 分段规划，特别是连续订购的模型 各个音素片段，使用现有的数据进行评估 自发言语中的错误语料库，并辅以额外的 需要收集的例子。 前景规划模型，特别是 语调突出（音调重音）的位置，在 实验得出的文本句子以及数据库中的 带有视觉标记的调频广播演讲。 最后给出一个模型 分段规划和前景规划的整合通过来自以下数据的测试： 分段错误的前景分析，通过收集 来自自发言语和来自的录音错误的新语料库 通过实验引出旨在对比效果的话语 错误模式上的句法结构与意象结构。 在此过程中 工作，现有的言语错误数据库和带有 Prosopically 标签的数据库 演讲内容将被扩大并重新组织，以便于 将它们分发给研究界。 对这些语音方面有更全面的了解 演讲制作计划流程，以及它们的关系 一方面是句法/语义规划，另一方面是发音运动 另一方面，控制将在几个领域提供优势，包括 为治疗学的发展奠定了完善的理论基础 针对弱势说话者（包括听力障碍者和 患有某种失语症的患者），为 自动语音识别和合成方面的进步将有助于 开发针对语言和言语障碍的辅助工具，以及 更有基础的方法来有效地教授韵律 英语作为第二语言。