Linguistics-Based Preference Information Modeling for Design Decision-Making

该奖项是根据2009年《美国复苏和再投资法案》(公法111-5)提供资金的。该研究奖的目的是对自然语言工程设计文本中嵌入的偏好信息进行建模，以确定偏好的语言形式，这些形式将构成支持计算决策与实际决策比较的决策模型的基础。产品设计过程的一种观点是，它是由设计师驱动的，他们对一系列可能的设计选择中的备选方案有偏好。这种偏好信息隐含在工程设计文本中，但可能很难从非结构化信息中提取。挑战在于对这些偏好进行语言学建模，并将它们映射到一个适合支持设计决策的数学模型中。这项工作将识别偏好的语言形式，产生一个全面的“偏好词典”，开发正式的偏好数学模型，并生成决策模型，以便将计算的决策与实际决策进行比较，以验证其有效性。如果成功，由于决策在工程设计过程中的基础作用，这项工作将影响许多行业，包括产品开发、汽车、航空航天和军事。这项研究旨在促进对设计语言的基本理解，特别是偏好是如何表达的。反过来，这将进一步加深对决策主观方面的基本知识，并朝着开发可用的、有效的决策支持方法的方向发展。其结果将是一种从非结构化设计文本中输入偏好信息以及决策信息的方法，该方法利用设计语言模型和概率提取。研究生将通过一门关于工程设计决策的跨学科、基于项目的课程来学习这项工作。一个不同的本科生群体将有实践的研究经验与这项工作，本科生的研究机会计划。

This award is funded under the American Recovery and Reinvestment Act of 2009 (Public Law 111-5). The objective of this research award is to model the preference information embedded in natural language engineering design texts in order to identify linguistic forms of preference that will form the basis for a decision-making model that supports comparison of computed decisions to actual decisions. One view of the product design process is that it is driven by designers who have preferences for alternatives within a set of possible design choices. Such preference information is implicit within engineering design texts, but can be difficult to extract from unstructured information. The challenge is in linguistically modeling these preferences and mapping them into a mathematical model suitable for supporting design decision-making. Work will identify linguistic forms of preference, produce a comprehensive 'preference lexicon', develop formal mathematical models of preferences, and generate a decision-making model so that computed decisions can be compared with actual decisions to verify their validity.If successful, this work will have impact across many industries, including product development, automotive, aerospace, and the military, due to the fundamental role of decision-making in the engineering design process. This research is intended to advance fundamental understanding of the language of design, in particular how preferences are expressed. In turn, this will further basic knowledge of the subjective aspects of decision-making and move towards the development of usable, effective decision support methods. The result will be an approach for imputing preference information as well as decision information from unstructured design texts that draws on both design language models and probabilistic extraction. Graduate students will learn about this work through an interdisciplinary, project-based class on decision-making in engineering design. A diverse group of undergraduates will have hands-on research experience with this work the Undergraduate Research Opportunities Program.