Workshop: Learning Hidden Linguistic Structure; January 3-7, 2019, New York New York

这次研讨会将汇聚来自语言学和自然语言处理(NLP)这两个基本上互不相关的科学界的主要语言研究人员。会议将创建一个关于面向语言的计算建模的智力交流论坛，并促进在语言学和自然语言规划社区之间建立富有成效的联系。为了最大限度地使更广泛的语言学界了解讲习班，并促进跨学科交流，讲习班将与美国语言学会年会同时举行，并将有来自NLP社区的知名特邀演讲者和小组成员参加。将语言学原理和人类语言学习的结果与统计机器学习的尖端计算方法相结合的跨学科研究有望带来互惠互利的科学突破，对研究界和整个社会都是如此。NLP的计算和数学建模方法的应用有望丰富对人类语言学习过程的科学理解，以及它如何依赖于儿童和成人语言环境中存在的信息。这种整合还有可能导致新的和改进的语言技术，如机器翻译和自动语音识别系统，它们通过促进不同语言使用者之间的交流，通过增加对网络上信息和教育资源的多语种获取，并通过为言语、听力和语言障碍者生产新的工具和资源，在现代社会发挥着越来越重要的作用。此外，讲习班能够接触到更广泛的语言学界，为从事语言学的妇女开辟了一条进入STEM的途径，这是不存在的，并提供与NLP社区的教育、培训和研究联系。讲习班的主题是“学习隐藏的语言结构”，之所以选择这一专题，是因为这两个社区在处理这一基本学习问题方面有强大但基本上不同的研究传统。隐含结构是语言学理论和人类语言学习理论的共同组成部分，但关于儿童如何从语言输入中习得这种表征，我们仍然知之甚少。另一方面，自然语言处理的研究已经产生了丰富的计算技术来模拟隐藏结构及其学习，但这些模型很少包含语言学原理。综合研究有可能导致语言技术的改进，特别是在资源匮乏的环境中，在那里，成功在很大程度上取决于从有限的数据中以适当的语言方式进行概括的能力。它还有可能在理解人类语言学习所依据的各种计算和隐藏的表征方面取得科学突破。研讨会将邀请受邀的演讲者从跨学科的角度介绍“学习隐藏的语言结构”这一主题。为了进一步加强语言学界和NLP社区之间的联系，还将与受邀的小组成员举行一次特别会议，主题为“语言学家应该对NLP有什么了解？”最后，为了促进语言学家在计算方法方面的培训，研讨会将主办两个关于NLP机器学习方法的入门级教程。这个奖项反映了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。

This workshop will bring together leading language researchers from two largely disjoint scientific communities: Linguistics and Natural Language Processing (NLP). The meeting will create a forum for intellectual exchange on linguistically oriented computational modeling and facilitate building of productive ties between the linguistics and NLP communities. To maximize the accessibility of the workshop to the broader linguistics community and to facilitate cross-disciplinary exchange, the workshop will take place in conjunction with the annual meeting of the Linguistic Society of America and will feature prominent invited speakers and panelists from the NLP community. Cross-disciplinary research that integrates principles of linguistics and findings from human language learning with cutting-edge computational methods from statistical machine learning promises to lead to scientific breakthroughs of mutual benefit to both research communities and to society more generally. Application of computational and mathematical modeling methods from NLP promises to enrich scientific understanding of the human language learning process and how it depends on the information present in children and adults' linguistic environments. The integration also has the potential to lead to new and improved language technologies, such as machine translation and automatic speech recognition systems that are playing an increasingly important role in modern society by facilitating communication between speakers of distinct languages, by increasing multilingual access to information and educational resources on the web, and by producing new tools and resources for people with speech, hearing, and language disabilities. In addition, the accessibility of the workshop to the broader linguistics community creates a pathway into STEM for women in linguistics that would not otherwise exist, providing educational, training and research connections with the NLP community.The theme for the workshop is "learning hidden linguistic structure", which is a topic chosen specifically because there are strong but largely separate research traditions approaching this fundamental learning problem in the two communities. Hidden structure is a common component of linguistic theories and of theories of human language learning, but much is still unknown about how such representations are learned by children from their linguistic input. On the other hand, NLP research has produced a wealth of computational techniques for modeling hidden structure and its learning, but these models rarely incorporate linguistic principles. Integrative research has the potential to lead to improvements in language technologies, especially in low-resource settings where success depends most on the capacity to generalize in linguistically appropriate ways from limited data. It also has potential to lead to scientific breakthroughs in understanding the sorts of computations and hidden representations that underlie human language learning. The workshop will host invited speakers who will present on the theme of "learning hidden linguistic structure" from interdisciplinary perspectives. To further increase ties between the linguistics and NLP communities, there will also be a special session with invited panelists on the topic of "What should linguists know about NLP? ". Finally, to facilitate the training of linguists in computational methods, the workshop will host two entry-level tutorials on machine learning methods in NLP.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.