A fair amount of work, such as map of science visualization methods, has been done in the area of visualization of scientific discovery and of the relationships among scientific disciplines. Typically, both the maps and the portfolio analyses are derived from keyword and/or citation analyses of research papers coupled with categorizations by discipline of journals and conferences. Although useful, this analysis is far from complete because it does not consider the full text of the papers, just the keywords and citations, and does not consider other sources, such as research project abstracts compiled by funding agencies and reports published by agencies or research organizations. A complete analysis upon which to base policy decisions, evaluations of the effectiveness of funding, or assessments of the direction of a field would include an integrated analysis of all these sources.
Intellectual Merit: This project develops a visual analytics approach to perform these assessments of multiple sources, including full text of papers, abstracts, and reports that have not been available before. The approach is exploratory, supporting investigations where one does not initially know precisely what one is looking for but rather uses tools that permit the discovery of new relations and the uncovering of insights. Once found these insights can be looked at in more detail, tested with the gathering of new evidence, and then be the basis of further insight discovery. To support this exploratory investigation, analyses must be, at least in initial stages, unstructured and automated. The most significant words and relations must bubble up from the texts themselves. They must be automated because there will be too many text documents to assess in any other way. Yet, the analyses cannot be completely automated; there must be a place to insert understanding, to organize and make sense of what is found and to direct the investigation in a new direction based on what is found. This is exactly where interactive visual analytics makes its contribution, revealing to investigators detailed results in understandable visual displays, providing clues to prompt further exploration, and supporting organization and annotation of collected evidence and pursuit of new hypotheses. This project also looks at changes and trends over time in the paper and other collections. Detailed examination of changes and trends over time brings out behaviors that may be caused by new or newly revealed directions chosen by researchers in a field, by changes in funding, or by new and important applications. The approach that is applied is based on analyses of streaming text organized into stories, reports, or similar narrative structures. The streaming stories are organized on the fly into ?event clusters? of similar stories that begin and end at particular points in time and have detailed time structures. The goal is to identify motivating events such as new funding directions, new directions established by leaders in a field, as well as new interdisciplinary thrusts across fields.
Broader Impacts. As indicated by the recent Visualization of Scientific Discovery Workshop (September, 2008) with sponsorship by NSF, the DOE Office of Science, collaboration by several NSF divisions, and broad attendance by program managers, researchers, and business innovators, there is a significant interest and need for visualization. In addition, the just-released Science of Science Policy Roadmap (November, 2008) from the Office of Science and Technology Policy and the National Science and Technology Council prominently mentions the need for visualization, and in particular visual analytics, tools for science analysis. This report also mentions the need for assessment of these tools for real science analysis applications. This project is positioned to meet both these needs by pursuing the development and assessment of a broad, flexible visual analytics approach for real science and innovation policy applications with real data.
William Ribarsky, Wenwen Dou, and their team developed tools for examining the relationship between science funding and the evolution of research fields. These tools use a visual analytics approach. The tools makes it possible to analyze large number of research proposals over a long time period in order to trace the progression of activity in a particular area. With these tools, Dr. Ribarsky explained, it is possible to analyze automatically and then represent visually the prevalence of particular ideas in proposals, papers, and the broader media. The visual representations can reveal trends, the impact of events or relationships, and the possible cause and effect relationships. The central tool (HierarchicalTopics) can be used for general exploration and to identify general trends or to conduct more detailed analyses. An example of the information that the visual representation can summarize is provided in Figures 1 and 2. This HierarchicalTopics visualization traces trends in proposals awarded by the NSF Computer and Information Science and Engineering Directorate from 2005 to 2012. The data shown include 11,961 awarded proposals. Two rising research areas are shown in Figure 2, with the contributing programs identified as the possible cause for the increase. These and other trends depicted were confirmed through evaluations by a former program manager for the directorate. Case Studies: 1. Depicting temporal portfolio of NSF programs Using the HierarchicalTopics tool, a domain expert started by visually browsing all hierarchical topic groups that are produced by the TRT algorithm. He quickly identified a few topics of interest and interactively merged them into topic groups that fit his analytic goal. The result of his customized grouping and corresponding annotation is shown in the first column in Figure 1. Specifically, two groups of topics are created through the "join" and "collapse" interactions, "HCI" and "Information Retrieval and Data Mining (IR)". With the exploration scope narrowed down to these two topic groups, the user wanted to identify and compare the trends in research funding for individual group over the years. Therefore, he turned to the Hierarchical ThemeRiver view and selected the two topic groups for closer examination. The second column in Figure 1 illustrates the overall temporal evolution of selected groups. The user noticed that the trend of proposals awarded under the IR group seemed steady with a slight decline over the recent two years. To examine and compare the development of individual topics in the IR group, the users further isolated three topics that are of interest. The corresponding trends for these topics are shown next to the overall trend. Through quickly examining the volume of each topic trend, the user confirmed his hypothesis that topic 18 on "web search and document retrieval" has continued to be a more popular subfield over the years in terms of NSF funding. However, the user was surprised when he found that the "HCI" group exhibits a slight decline in recent years after a steady growth around 2007. Through examining individual topic trends, more interesting patterns emerged. Although the overall trends for other topics group have subsided slightly, the research on "affective computing and emotion related studies" has gone up significantly in the past two years, as outlined in Orange. This use case illustrates that the visual interface not only enables the user to view trends for topics that describe a research field, but also permits the discovery of the contributions of individual topics to the overall trends as well as anomalies. According to the user, such analysis gave him valuable insights in understanding the research trends in the areas of his interest and could potentially help him adjust future proposal focus. 2. Identifying program impacts in research Given that the two topic groups exhibit slightly downward trending, the user wanted to identify research topics that received more funding interest in recent years. He started by mouse hovering over each topic ribbon in the main Hierarchical ThemeRiver view, looking for increasing trends. Two topic groups caught his attention as shown in Figure 2. Both groups exhibit increasing volume in the past three years, indicating more research proposals were awarded in the two areas. The top row illustrates a topic group related to environmental research as well as citizen science. As shown in the individual temporal trend for each topic, the user identified that the topic on citizen science and spatial temporal analysis significantly contributed to the recent growth of the focused topic group. The second row in Figure 2 illustrates a topic group that summarizes research on medical and healthcare related research. Through enabling the time column selection, the user selected proposals related to the health care topic that were awarded in 2012, highlighted in the yellow rectangle. He then discovered that most of the proposals were related to health monitoring and awarded by the only-recently launched program–Smart and Connected Health (2011).
