Requirements for domain-specific data access of long-term interaction data in smart environments
Requirements for domain-specific data access of long-term interaction data in smart environments
复制标题
智能环境中长期交互数据的特定领域数据访问要求
DOI：
--
复制
发表时间：
2015
期刊：
影响因子：
--
通讯作者：
P. Cimiano
中科院分区：
文献类型：
--
作者：
N. Köster;S. Wrede;P. Cimiano
近年来，对交互环境中的人工认知和智能系统的研究进一步扩展到支持长期场景的更复杂的应用[1]。随着传感器和致动器数量的增加，这些应用导致以快速的时间尺度从各种模态（例如语音、对话、人或情况）产生大量提取的数据。如何支持存储和访问在传感器密集型环境中长时间生成的大量数据的问题迄今尚未得到足够的关注。许多现有的方法具有以机器人为中心的方法，因此数据存储和访问主要集中在影响机器人动作和行为的方面。因此，我们认为，存储和访问数据的人机交互（HRI）域需要简化和扩展的开发人员。中心研究问题包括HRI相关概念的可访问性，在线查询评估的可能性，以及将机器学习纳入数据存储和检索。为了解决这些差距，我们考虑了一种新的方法，其目标是在很长一段时间内改善交互相关知识的存储和访问，重点是捕获相关数据和概念，同时使其在线可查询。因此，我们建议开发一个适当的抽象领域特定的数据访问，封装和简化检索这些数据的交互相关组件的开发人员。在下文中，我们希望进一步分析该领域，并提取对数据结构、表示、存储、粒度、存储策略等的第一个需求，以促进此类功能。
In recent years research on artificial cognitive and intelligent systems in interactive environments expanded further to more complex applications that support long-term scenarios [1]. With the increasing amount of sensors and actuators, these applications lead to the production of vast amounts of extracted data from various modalities (e.g. speech, dialog, persons, or situations) at rapid time scales. The issue of how to support storage and access to large amounts of data as generated in sensor-intensive environments over longer periods of time has not received sufficient attention so far. Many existing approaches have a robot-centric approach and thus data storage and access focuses primarily on aspects impacting the robot actions and behaviours. We hence argue that storage and access to data about the human robot interaction (HRI) domain needs to be simplified and extended for developers. Among central research questions are the accessibility of HRI related concepts, the possibility for online query evaluation, and the incorporation of machine learning in data storage and retrieval. Addressing these gaps, we consider a new approach that targets to improve storage and access of interaction related knowledge over long periods of time, focusing on capturing relevant data and concepts whilst making it online queryable. We thus propose the development of an appropriate abstraction for a domain-specific data access that encapsulates and simplifies the retrieval of such data for developers of interaction relevant components. In the following, we want to analyse the domain further and extract first requirements for data structures, representations, storage, granularity, storage policies, etc. to facilitate such functionality.
DOI：
10.1109/iros.2014.6942700
发表时间：
2014-09
期刊：
2014 IEEE/RSJ International Conference on Intelligent Robots and Systems
影响因子：
--
作者：
André Dietrich;Sebastian Zug;Siba Mohammad;J. Kaiser
通讯作者：
André Dietrich;Sebastian Zug;Siba Mohammad;J. Kaiser