Highly Interactive Visual Analytics

视觉分析(VA)被定义为通过交互式视觉界面促进分析推理的科学。与单纯的计算相比，可视化可以利用人类大脑与生俱来的视觉智能来获得对复杂数据的新见解，这些复杂数据可能包含事实上的不确定性，或者发生的时间和地点。虽然创建图形可视化的方法是理解的，但视觉分析(VA)直接解决了这一过程的最终目标：人类对复杂情况的推理和分析。 在这个提案中，我的目标是开发一个新的可视化分析软件体系结构模型，该模型更直接地考虑到分析人员在分析大量文档期间所从事的各种认知活动。目前，审查大型数据集合的既定做法是使用信息检索工具来分析文档集的趋势，并搜索支持或不支持从审查文件中可能产生的假设的证据。信息可视化工具，如节点链接图，用于绘制文档中实体之间的连接，并使用时间编码事件绘制地图，以可视化人员在城市中的移动，仅举几个例子。 这些视觉工具的优势在于它们提供了一种外部认知的手段：它们用感兴趣的关系的具体空间表示来增强内部认知过程，稍后可以参考这些空间表示来进行进一步的认知活动。目前的视觉工具的缺点是，它们本身并不能为更高层次的认知活动提供更多的直接支持。例如，许多分析工作需要追求一些非正式假设，这些假设必须进行测试，并与其他可能的假设进行比较，以确定是否成立。此外，许多重要的分析任务需要很长一段时间才能完成，必须由许多人分担。在我与情报分析员的讨论中，经常有人呼吁建立能够概括过去分析要点的系统，并对新的或更新的数据重复使用分析步骤。 这项提案的重点是设计和开发新的模型和软件，以协助分析过程。这项工作的第一个新要素是为可视分析提供一个可编辑和可管理的规范系统，该系统除了显示文档集合和其中的实体的多种可视表示外，还可以管理分析过程。其次，这个系统将允许分析员与他们的同事比较、分析和分享分析策略，从而能够提供多层次的分析来源。

Visual Analytics (VA) is defined as the science of analytical reasoning facilitated by interactive visual interfaces. In contrast to computation alone, visualization can harness the human mind's innate visual intelligence to gain novel insights into situations characterized by complex data that may contain uncertainty in fact, or time and location of occurrence. While methods for creating graphical visualizations are understood, visual analytics (VA) directly addresses the ultimate goal of this process: human reasoning and analysis about complex situations. In this proposal, I aim to develop a new architectural model of Visual Analytics software that more directly takes into account the various cognitive activities that analysts engage in during the analysis of large collections of documents. Currently, established practice for examining large data collections is to use information retrieval tools to analyze a document set for trends and to search for evidence that supports or does not support hypotheses that may arise from examining the documents. Information visualization tools such as node-link diagrams are used to plot connections between entities in documents, and maps are plotted with time-coded events to visualize the movement of people across a city, to name a few examples. The strength of these visual tools is that they supply a means of external cognition: they augment internal cognitive processes with concrete spatial representations of relationships of interest that can be referred to later for further cognitive activity. The drawback of current visual tools is that they do not themselves offer much direct support for more "high level" cognitive activity. For example, much analytical work requires the pursuit of a number of informal hypotheses that must be tested and compared with other possible hypotheses for confirmation or disconfirmation. Moreover, many significant analysis tasks take place over a long period of time and must be shared among many people. In my discussions with intelligence analysts, there has been a frequent call for systems that are capable of recapitulating key points of past analyses, and for the re-use of analytical steps on new or updated data. This proposal focuses on the design and development of new models and software to aid in the analysis process. The first new element of this work is to provide an editable and manageable specification system for visual analysis that can manage the analytical process in addition to displaying multiple visual representations of document collections and the entities within them. Second, this system will allow analysts to compare, analyze and share analytical strategies with their colleagues, to enable the delivery of multiple levels of analytical provenance.