III: Small: Collaborative Research: Learning to Model Sequences

III：小：协作研究：学习序列建模

• 批准号: 1217686
• 负责人: Thorsten Joachims
• 金额: $ 31.4万
• 依托单位: Cornell University
• 依托单位国家: 美国
• 项目类别: Continuing Grant
• 财政年份: 2012
• 资助国家: 美国
• 起止时间: 2012-10-01 至 2016-09-30
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=1217686&HistoricalAwards=false
• 关键词: III; Small; Collaborative; Research; Learning

This collaborative research project involves faculty and students from Cornell University (IIS-1217686) and Ithaca College (IIS-1217485) in an interdisciplinary project. The ability to learn predictive models of sequences is a component in several application problems, ranging from language models for machine translation to the recommendation of material to read in order to master a subject. The goal of this project is to develop new machine learning algorithms that can learn sequence models for items that are difficult to describe by attributes. In particular, the project develops models that automatically embed items in a latent feature space based on training sequences, that can integrate partial and noisy side information, and that have the ability to model long-range dependencies.The resulting predictive models have a potential to be employed in science and education and can support the economic shift towards online business applications. The project focuses on the recommendation of music playlists as the main testbed. A deployed online music recommendation system not only provides the framework for testing and evaluation, this application domain helps to attract a broad spectrum of students from collaborating institutions, Cornell University and Ithaca College (an undergraduate institution), enabling the integration of undergraduates in the research. Project results, including open source software and annotated data set, are disseminated via the project Web site (http://www.cs.cornell.edu/People/tj/playlists/).

该合作研究项目涉及康奈尔大学（IIS-1217686）和伊萨卡学院（IIS-1217485）的教师和学生参与一个跨学科项目。学习序列预测模型的能力是许多应用问题的一个组成部分，从机器翻译的语言模型到为掌握一门学科而推荐阅读的材料。该项目的目标是开发新的机器学习算法，可以学习难以用属性描述的项目的序列模型。特别是，该项目开发了基于训练序列自动将项目嵌入潜在特征空间的模型，可以集成部分和噪声侧信息，并且具有建模远程依赖关系的能力。由此产生的预测模型有可能被应用于科学和教育领域，并能支持经济向在线商业应用的转变。该项目将音乐播放列表推荐作为主要测试平台。一个部署的在线音乐推荐系统不仅提供了测试和评估的框架，这个应用领域有助于吸引来自合作机构的广泛的学生，康奈尔大学和伊萨卡学院（一个本科机构），使本科生能够整合到研究中。项目结果，包括开源软件和注释数据集，通过项目网站（http://www.cs.cornell.edu/People/tj/playlists/）传播。