Machine learning for discovering conversation strategies and recovering from breakdowns

用于发现对话策略并从故障中恢复的机器学习

• 批准号: 476952-2014
• 负责人: Rudzicz, Frank
• 金额: $ 1.82万
• 依托单位: University of Toronto
• 依托单位国家: 加拿大
• 项目类别: Engage Grants Program
• 财政年份: 2014
• 资助国家: 加拿大
• 起止时间: 2014-01-01 至 2015-12-31
• 项目状态: 已结题
• 来源: https://www.nserc-crsng.gc.ca/ase-oro/Details-Detailles_eng.asp?id=565514
• 关键词: Machine; learning; discovering; conversation; strategies

'Conversations with data' is a design metaphor behind IBM Watson Analytics, which offers a broad set of predictive and exploratory capabilities. Watson Analytics' ultimate objectives are to engage data scientists in increasingly intuitive and natural ways by making systems easier to use and smarter in its analyses. In this research, we are going to lay the groundwork for a speech-based conversation model trained in Watson Analytics capable of learning and detecting the context of analysis. This model will have a number of important characteristics: a) context awareness, b) the ability to handle communication breakdowns or uncertainty, and c) adaptability. For example, in context awareness, the system will recognize a query such as "my revenue for 2013 broken down by product and location" to indicate the beginning of a new analysis unit,

“与数据对话”是IBM沃森分析背后的一个设计隐喻，它提供了一系列广泛的预测和探索功能。沃森分析的最终目标是通过使系统更易于使用和更智能地进行分析，以越来越直观和自然的方式吸引数据科学家。在这项研究中，我们将为在沃森分析中训练的基于语音的对话模型奠定基础，该模型能够学习和检测分析的上下文。这个模型将具有许多重要的特征：a）上下文感知，B）处理通信故障或不确定性的能力，以及c）适应性。例如，在上下文感知中，系统将识别诸如“我的2013年收入按产品和位置细分”的查询，以指示新分析单元的开始，