WORKSHOP: Language Engineering Workshop for Students and Professionals Integrating Research and Education (WS98)

研讨会：为学生和专业人士提供研究与教育相结合的语言工程研讨会 (WS98)

• 批准号: 9732388
• 负责人: Frederick Jelinek
• 金额: $ 75万
• 依托单位: Johns Hopkins University
• 依托单位国家: 美国
• 项目类别: Standard Grant
• 财政年份: 1998
• 资助国家: 美国
• 起止时间: 1998-05-01 至 2000-04-30
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=9732388&HistoricalAwards=false
• 关键词: WORKSHOP; Language; Engineering; Workshop; Students

While spoken language systems are now used for many applications, language engineering technology is still not adequate for all important tasks. For instance, automatic recognition of natural conversational speech results in incorrect transcription of one- third of the words spoken. Parsing sentences from newspaper articles leads to faulty automatic analysis of half the sentences. The need to make progress in this important area is obvious. The number of available language engineering personnel is small and relatively few universities presently educate students capable of performing the required tasks. The Center for Language and Speech Processing at the Johns Hopkins University is organizing a six week workshop on Language Engineering from July 13 to August 24, 1998, in which teams of professionals and students will work together to advance the state of the art. Three topics will be addressed: (1) Dynamic Segmental Models of Speech Coarticulation; (2) Rapid Speech Recognizer Adaptation to New Speakers; (3) Core Natural Language Processing Technology Applicable to Multiple Languages. University professors, leading industrial and government researchers presently working in widely dispersed locations, graduate students, and six or more undergraduate students selected through a nationwide search will participate in this workshop. Significant technological innovations, training of students, cross-fertilization of ideas from various disciplines, initiation of collaborative research, and creation of valuable training and test corpora are expected to result from this endeavor.

虽然口语系统现在用于许多应用，但语言工程技术仍然不足以完成所有重要任务。例如，对自然会话语音的自动识别会导致三分之一的单词转录错误。从报纸文章中分析句子会导致对一半句子的错误自动分析。在这一重要领域取得进展的必要性是显而易见的。可用的语言工程人员的数量很少，目前相对较少的大学培养能够执行所需任务的学生。约翰霍普金斯大学的语言和语音处理中心将于1998年7月13日至8月24日组织一个为期六周的语言工程研讨会，由专业人员和学生组成的团队将共同努力，推动语言技术的发展。将讨论三个主题：(1)语音协同发音的动态片段模型；(2)语音识别器对新说话者的快速适应；(3)适用于多语言的核心自然语言处理技术。参加此次研讨会的有大学教授、目前在世界各地工作的主要产业界和政府研究人员、研究生以及通过全国范围内选拔的6名以上的本科生。重要的技术革新、对学生的训练、不同学科思想的交流、合作研究的启动、以及有价值的训练和测试语料库的创建都有望从这一努力中得到结果。