IRES: Avatar-based Adaptive Context System

IRES：基于阿凡达的自适应上下文系统

• 批准号: 1458272
• 负责人: Avelino Gonzalez
• 金额: $ 23.2万
• 依托单位: The University of Central Florida Board of Trustees
• 依托单位国家: 美国
• 项目类别: Standard Grant
• 财政年份: 2015
• 资助国家: 美国
• 起止时间: 2015-06-15 至 2020-05-31
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=1458272&HistoricalAwards=false
• 关键词: IRES; Avatar; based; Adaptive; Context

Part 1This interdisciplinary partnership between the University of Central Florida and the Fraunhofer Institute for Digital Media Technology (FIDMT) in Ilmenau, Germany is focused on adaptive conversational avatars, the rapidly emerging field crossing computer engineering, computer science, education, communications, and social science. Immediate applications of this research field include artificial intellegence and national security (including cyber-security), interactive robotics, improvement of quality of life for disbaled, and health and caretaking for children and elderly.This project will place students from the University of Central Florida under the mentorship of the PI (Dr. A. J. Gonzalez) and of Dr. Klaus Jantke, the counterpart at FIDMT in Ilmenau, Germany. Dr. Jantke is the director of the Children's Media Department of FIDMT located in Erfurt, Germany and has a long and illustrious history in research in computing media. The international aspect of innovative and advanced research is essential in modern research hence the PIs will work with three cohorts of students, one during each year of the project's existence. Each cohort will include one graduate student and either two undergraduates (the first year) or four (in each of the subsequent years). The research period for each cohort will be 16 weeks - eight weeks in the US and eight weeks in Germany for each year of the grant period.This research project is motivated by an ancient art of storytelling. In our pursuit of an artificially intelligent computer agent, the IRES project seeks to build a capability to autonomously synthesize possible scenarios for the system development and to modify them dynamically upon listener request. More specifically, the topic of the research in this project is the creation of an avatar-based system that can synthesize and adapt a scenario according to the user's request in real time and without any pre-scripted pathways. Good storytellers were treasured in medieval times, given the lack of other media through which to relate a story to a mostly illiterate population. Therefore, the project seeks to embody the storyteller in a lifelike avatar that resembles an actual person. This avatar will tell the story to the listener in spoken natural language, and interact with her/him when the latter requests changes to the story. Part 2Storytelling media have evolved over time, from oral stories to modern E-books. Since the development of the computer, storytelling systems have become a science of their own, and have evolved from simple systems that can only generate a single short story to systems that respond to the listener's actions by modifying the story dynamically in real time. Digital storytelling has therefore become a growing field within artificial intelligence. The project seeks to take this evolution of storytelling media one step further by doing research to create a virtual storyteller who tells a dynamic story. The story is modifiable through a request by the listener (typically a child, a student, or an elderly person), yet will seek to remain realistic as well as interesting. Every story has a story space. That is, only so many things can happen in a story. We use contextual reasoning to represent the story space. In the real world, courses of action are influenced by the current context, making some conversational avatars very attractive while others unattractive when addressing the current situation within the story space. In a similar manner, the situation faced by the protagonist in the dynamic scenario will limit the choices of actions that he/she would otherwise have, thereby taking the story in various directions, none of which need be specifically pre-scripted. The PIs base the proposed research on the use of formal methods to manipulate the story space within the main theme of the story. By formal methods the PIs mean that one represents the story knowledge formally in terms of strings, interaction sequences such as storyboards and graphs, formulas (for conditions), and the like. Formal methods, therefore, will give the ability to reason with formal methods (string comparison, unification, anti-unification and the like) in the story space. Formal methods have been used in the literature to manipulate contextual information.

Part 1中央佛罗里达大学和位于德国伊尔梅瑙的弗劳恩霍夫数字媒体技术研究所（Fraunhofer Institute for Digital Media Technology，简称FDMT）之间的跨学科合作关系主要集中在自适应会话化身，这是一个跨越计算机工程、计算机科学、教育、通信和社会科学的新兴领域。 该研究领域的直接应用包括人工智能和国家安全（包括网络安全），交互式机器人，改善残疾人的生活质量，以及儿童和老人的健康和照顾。该项目将在PI（A. J. Gonzalez）和德国伊尔梅瑙FIDMT的同行Klaus Jantke博士。Jantke博士是位于德国埃尔富特的FIDMT儿童媒体部主任，在计算机媒体研究方面有着悠久而辉煌的历史。创新和先进研究的国际方面在现代研究中是必不可少的，因此PI将与三组学生合作，每年一次。每个队列将包括一名研究生和两名本科生（第一年）或四名（随后的每一年）。 每个队列的研究期为16周--在美国为8周，在德国为8周，每年的资助期。这个研究项目的动机是一种古老的讲故事的艺术。在我们追求的人工智能计算机代理，IRES项目旨在建立一个能力，自主合成可能的情况下，系统的发展，并根据听众的请求动态地修改它们。更具体地说，该项目的研究主题是创建一个基于化身的系统，该系统可以根据用户的请求在真实的时间内合成和调整场景，而无需任何预先编写的路径。在中世纪，优秀的说书人受到珍视，因为缺乏其他媒体来将故事与大多数文盲人口联系起来。因此，该项目试图将讲故事的人体现在一个逼真的化身中，类似于一个真实的人。这个化身将以自然语言向听众讲述故事，并在后者请求改变故事时与她/他交互。第二部分讲故事的媒体随着时间的推移而发展，从口头故事到现代电子书。自从计算机发展以来，讲故事系统已经成为一门独立的科学，并且已经从只能生成单个简短故事的简单系统发展到通过在真实的时间中动态地修改故事来响应听众的动作的系统。因此，数字讲故事已成为人工智能中一个不断增长的领域。该项目旨在通过研究创造一个讲述动态故事的虚拟讲故事者，进一步推动讲故事媒体的发展。故事可以根据听众（通常是孩子、学生或老人）的要求进行修改，但要保持现实和有趣。每个故事都有一个故事空间。也就是说，只有这么多的事情可以发生在一个故事。我们使用上下文推理来表示故事空间。在真实的世界中，行动过程受到当前上下文的影响，使得当在故事空间内解决当前情况时，一些对话化身非常有吸引力，而另一些则没有吸引力。以类似的方式，在动态场景中主角所面临的情况将限制他/她原本具有的行动的选择，从而将故事带到各个方向，其中没有一个需要特别预先编写。 PI的基础上提出的研究使用正式的方法来操纵故事的主题内的故事空间。通过形式化方法，PI意味着人们用字符串、交互序列（如故事板和图形）、公式（用于条件）等形式化地表示故事知识。因此，形式方法将赋予在故事空间中使用形式方法（字符串比较、统一、反统一等）进行推理的能力。在文献中已经使用了形式化的方法来操纵上下文信息。