Optimizing Language Learning through Content-Driven Machine Learning

通过内容驱动的机器学习优化语言学习

• 批准号: RGPIN-2020-04727
• 负责人: Johns, Brendan
• 金额: $ 3.28万
• 依托单位: McGill University
• 依托单位国家: 加拿大
• 项目类别: Discovery Grants Program - Individual
• 财政年份: 2021
• 资助国家: 加拿大
• 起止时间: 2021-01-01 至 2022-12-31
• 项目状态: 已结题
• 来源: https://www.nserc-crsng.gc.ca/ase-oro/Details-Detailles_eng.asp?id=740149
• 关键词: Optimizing; Language; Learning; through; Content

Lexical experience is a fundamental organizer of the language processing system, where it is used in differing aspects of lexical cognition, from the organization of the mental lexicon to the acquisition of knowledge. Recent research within machine learning and computational cognitive science has led to the development of a class of models entitled distributional models of lexical semantics, which are designed to exploit statistical regularities within natural language to learn the meaning of words. These models have had remarkable success in accounting for language-based behaviours, from standard psycholinguistic experimental tasks to complex applied problems. However, there is one main failing of this approach: they do not take into account individual variability in learning history. Everyone has had different experiences with the world, and this leads to variability in language usage and comprehension. To overcome this problem, I developed a new machine learning framework, entitled experiential optimization, that can estimate the types of linguistic experience that an individual person might have had. The short-term goals of my research program is to gain a better understanding of the connection between lexical cognition and experience. In order to accomplish this goal, the proposed research uses a combination of theoretically-driven behavioural experimentation and large-scale computational modeling to further develop experiential optimization. Specifically, there are four points of knowledge that this grant proposes to discover: (a) a determination as to how past lexical experience impacts current lexical processing at a fine-grained level, (b) further develop the machinations of experiential optimization by using cutting edge methodology from machine learning, (c) generate a new, general test of reading experience, and (d) use the combined empirical and technological improvements to optimize an individual's learning potential. The results of these findings will lead to an improved empirical understanding of the connection between cognition and experience and also the development of corresponding cognitive technologies that can quantify this connection. The long-term goals of my research program is to use the cognitively-plausible machine learning and natural language processing models coming out of my basic research to develop automated educational technologies. The outcome of this grant will enable this through the continued development and refinement of new paradigms within cognitive computing and machine learning, while positioning Canada to continue to be a world leader in these areas. Students trained under this grant will receive state-of-the-art training in machine learning, cognitive modeling, data science, and psycholinguistics, a unique combination that will provide the trainees with valuable research experience that is desirable within both academia and industry.

词汇经验是语言处理系统的基本组织者，它被用于词汇认知的不同方面，从组织心理词汇到获取知识。最近在机器学习和计算认知科学领域的研究导致了一类名为词汇语义的分布模型的发展，该模型旨在利用自然语言中的统计规律来学习单词的意义。这些模型在解释基于语言的行为方面取得了显著的成功，从标准的心理语言学实验任务到复杂的应用问题。然而，这种方法有一个主要缺陷：它们没有考虑到学习历史中的个体差异。每个人对这个世界都有不同的经历，这导致了语言使用和理解的不同。为了克服这个问题，我开发了一个新的机器学习框架，名为经验优化，它可以估计个人可能拥有的语言经验的类型。我的研究项目的短期目标是更好地理解词汇认知和经验之间的联系。为了实现这一目标，本研究采用理论驱动的行为实验和大规模的计算建模相结合的方法，进一步发展经验优化。具体地说，这笔赠款打算发现四个知识点：(A)确定过去的词汇经验如何影响当前细粒度的词汇处理；(B)通过使用机器学习的尖端方法，进一步开发经验优化的机制；(C)生成一个新的、一般的阅读体验测试；以及(D)使用经验和技术改进相结合的方法来优化个人的学习潜力。这些发现的结果将导致对认知和经验之间联系的经验理解的改善，以及可以量化这种联系的相应认知技术的发展。我的研究计划的长期目标是使用我的基础研究得出的在认知上合理的机器学习和自然语言处理模型来开发自动化教育技术。这笔赠款的结果将通过继续开发和完善认知计算和机器学习领域的新范式来实现这一点，同时使加拿大在这些领域继续保持世界领先地位。根据这项资助培训的学生将接受最先进的机器学习、认知建模、数据科学和心理语言学方面的培训，这是一个独特的组合，将为学员提供学术界和工业界都希望获得的宝贵研究经验。