The webLyzard information exploration and retrieval interface (=”dashboard”) helps identify, track and analyze topics across stakeholders and sources. It provides access to a comprehensive content repository structured along geospatial, semantic and temporal dimensions. The dashboard is divided into six main content areas:
- Sources and Configuration. The upper menu contains a search field and specifies time interval, sources to be analyzed (news media, social media, etc.), and interface options such as language and sentiment filter (positive, neutral, negative). It also contains a menu item to export search results and other datasets in various formats.
- Topics. The upper left window of the dashboard provides access to either the topic management view (to define and compare topics) or the drill down view (to explore annotations such as sentiment related to the current search). Users have the following options:
(i) click on a label to trigger a search for a topic or filter by annotation;
(ii) activate markers (= small rectangles) to select topics or annotations to be shown in the charts;
(iii) select the first five items (or deselect all) with the ‘chart’ symbol next to a category name;
(iv) use the ‘settings’ symbol to use logical operators – replace, restrict, extend, exclude – to combine multiple topics or filter for specific annotations (e.g., positive sentiment); in the topic management view, the ‘settings’ symbol also allows configuring or deleting topics, setting email alerts, and accessing the topic’s Wiki page,
- Trend Charts. Interactive trend charts show (i) the share of voice, a comparative measure of attention based on the relative number of mentions, (ii) the frequency of selected topics in the specified time interval, (iii) the average sentiment regarding these topics, and (iv) the level of disagreement, measuring how strongly the sentiment regarding this topic fluctuates. Setting a moving average dampens the impact of short-term fluctuations and highlights longer-term trends.
- Content View. The section below the trend chart offers various ways to show the search results: documents, sentences, word tree, entities, relation tracker, sources and source map, as well as network graph. Interactive features in the ‘document’ and ‘sentence’ modes include: (i) mouse-over to preview documents; (ii) an initial click to select a document and show an extended quote; (iii) a second click to display the full text, which reveals the document’s annotations including keywords, location, sentiment, and relevance.
- Associated Terms. The lower left view of the dashboard displays a list of top associations with the search term, based on the selected source and time interval. Similar to the topic management section, the associated terms can be shown in the trend chart and used to trigger full-text searches.
- Maps and Visual Analytics. Visual means to investigate topic associations are the tag cloud (alphabetical structure) and the keyword graph (hierarchical structure). Additional visualizations include the geographic map to render the regional distribution of search results and the cluster map to show emerging stories (= sets of related documents). The maps are synchronized and can be re-positioned using drag-and-drop operations. The ‘pause’ button in the upper right corner disables dynamic updates (e.g. automated zoom in the geographic map, or the node positioning in the keyword graph).
Users can adjust the time interval (default: two months) and access historic data using the date selector. This is a global setting that not only affects the trend chart, but also limits search queries and dynamic visualizations to the chosen time interval. The interval can be specified by choosing a “from” and “to” date, or selecting the past [n] weeks or months.
Visual Tools and Adaptive Tooltips
The visualizations on the right side of the dashboard help to explore the underlying knowledge base. Clicking on the ‘maximize’ button increases the size of a map and adapts the displayed content. Clicking on the ‘popup’ button opens the map in a separate browser window and allows using the system in multiple-screen configurations. Users can switch maps on and off via the “Visualizations” menu, and re-arrange the layout by dragging them to the desired position.
User actions in one window trigger an immediate update of all other windows. Instead of entering query terms or using predefined topics to trigger a search, for example, the visualizations provide adaptive tooltips with logical operators to Replace, Restrict or Extend the current search, or Exclude certain terms from the search results. A small line chart previews the results of these modified searches. For some of the visualizations, the tooltip contains specific actions to filter the data or modify the display.
Usability and User Experience
To gain additional insight into the user experience of the dashboard and the perception of the various visual tools, we conduct usability evaluations in regular intervals. The aim is to determine strengths and weaknesses of the interaction design. In experimental settings, the gaze data of test users is recorded and analyzed with eye tracking software to generate heat maps and show which dashboard elements are used to complete a given task, and in which sequence.
- Scharl, A., Herring, D., Rafelsberger, W., Hubmann-Haidvogel, A., Kamolov, R., Fischl, D., Föls, M. and Weichselbraun, A. (2017). Semantic Systems and Visual Tools to Support Environmental Communication”, IEEE Systems Journal: Forthcoming.
- Scharl, A., Hubmann-Haidvogel, A., Sabou, M., Weichselbraun, A. and Lang, H.-P. (2013). From Web Intelligence to Knowledge Co-Creation – A Platform to Analyze and Support Stakeholder Communication, IEEE Internet Computing, 17(5): 21-29.
- Hubmann-Haidvogel, A., Scharl, A. and Weichselbraun, A. (2009). Multiple Coordinated Views for Searching and Navigating Web Content Repositories, Information Sciences, 179(12): 1813-1821.